<%BANNER%>

Simulation of Spatial and Temporal Variability of Methylmercury Concentration within Channel Banks and Surface Waters of...


PAGE 1

SIMULATION OF SPATIAL AND TEMPORAL VARIABILITY OF METHYLMERCURY CONCENTRATION WITHIN CHANNEL BANKS AND SURFACE WATERS OF THE CARSON RIVER, NEVADA By DANIEL E. CRAWFORD A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2003

PAGE 2

Copyright 2003 by Daniel E. Crawford

PAGE 3

This thesis is dedicated to the graduate students of the University of Florida.

PAGE 4

iv ACKNOWLEDGMENTS I would like to thank the Na tional Science Foundation for their generous support of this project through grant # 4510586-12. I wish to acknowledge the time, effort, and continued support of my adviso r John J. Warwick. The well-be ing of his students always remained a priority. Special thanks go to Andy James for his support and guidance throughout this research, most notably during th e 7 months after John’s departure. I also thank Bill Wise and Jean-Claude Bonzongo for serving on my committee and providing invaluable feedback. I thank my peers a nd closest friends for helping me (through countless hours of frustration) with words of encouragement; and for always being there. Lastly, I would like to thank my parents and family. Although the re search may not have been their forte, their support, encouragemen t, and sense of pride in my accomplishments never waned. My time at this university ha s been long-lived, but the experience was greater than ever expected and has tr uly prepared me for the future.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT.....................................................................................................................xi ii CHAPTER 1 RESEARCH LOCATION: AN INTROD UCTION TO THE CARSON RIVER........1 Carson River Basin.......................................................................................................1 Carson River Streamflow.............................................................................................2 Carson River Aquifer Systems.....................................................................................5 Historical Mining Operations.......................................................................................5 Carson River Geomorphology......................................................................................7 2 MERCURY CONTAMINATION AND MERCURY CYCLING............................10 Mercury Cycling within Surface Water Systems.......................................................10 Quantification of Mercury Levels in the Carson River System.................................11 Introduction of Mercury to the Carson River System................................................13 Chemistry of Mercury in Soil.....................................................................................15 Mercury Transformation and Toxicity.......................................................................16 Methylmercury in the Water Column.........................................................................19 Mercury in the Aquatic Food Chain...........................................................................23 Carson River Superfund Site......................................................................................27 Health Effects of Mercury Contamination.................................................................28 3 PREVIOUS MODEL STRUCTURE.........................................................................30 Hydrodynamic Modeling Using RIVMOD................................................................30 The WASP5 and MERC4 Programs...........................................................................33 Stages of Model Development....................................................................................36

PAGE 6

vi 4 CHANNEL BANK ELEMENTS: LINKING SURFACE WATER AND GROUNDWATER.....................................................................................................38 Modeling the Carson River.........................................................................................38 Available Data............................................................................................................40 Introduction to the Hyporheic Zone...........................................................................41 Modeling Surface-Subsurface Interac tions: A Historical Perspective.......................43 Modeling Simplifications...........................................................................................50 Using IFMTFM Modifications to Tr ack Bank Saturation: Bank Element Functionality..........................................................................................................50 Final Considerations and Comparative Analysis........................................................60 5 SURFACE WATER MODELING.............................................................................63 Site Selection..............................................................................................................63 Collection of GPS Survey Data..................................................................................64 Surface Water Model Selection..................................................................................66 Functionality of the FESWMS Model........................................................................67 Creation of Scatter Point Data Set..............................................................................70 Pre-Processing of Data Prio r to Mesh Construction...................................................70 Mesh Construction and Editing as Discussed in Appendix A....................................72 Summary of Mesh Editing..........................................................................................74 6 WATER TABLE RESPONSE TO STREAMFLOW VARIABILITY......................77 Linking Surface Water Modeli ng and Groundwater Data..........................................77 Groundwater Well Data..............................................................................................85 Quantifying the Relationship between Rive r Stage and Water Table Elevation........95 7 BANK ELEMENT DEVELOPMENT AND COMPARISON TO PREVIOUS IFMTFM...................................................................................................................137 General Considerations for Bank Element Design and Operations..........................137 Updated Input Data Sets to Include Bank Elements.................................................139 Comparison of Updated IFMTFM Code versus Previous Version..........................144 Next Step for the IFMTFM Model...........................................................................164 8 COMPARATIVE ANALYSIS: RIVMOD VERSUS FESWMS MODULE OF SMS..........................................................................................................................16 7 Comparison of Channel Geometry along the Well Transect....................................169 Steady-State Comparisons for Ve locity and Depths of Flow...................................169 Comparative Analysis and Data Limitations............................................................176

PAGE 7

vii APPENDIX A SMS Model Parameterization and Mesh Editing.....................................................179 B Water Table Response to Streamflow Oscillations..................................................239 C Element Drying and Water Surface Lowering within Completed SMS Model.......253 LIST OF REFERENCES.................................................................................................258 BIOGRAPHICAL SKETCH...........................................................................................266

PAGE 8

viii LIST OF TABLES Table page 4-1. Spatial variability of reac tion constants within IFMTFM........................................53 6-1. Classification of surveyed water surf ace elevations according to time of collection................................................................................................................109 6-2. Steady-state output for depth and discharge from final SMS model.....................124 7-1. Spatial variability of M/D r eaction constants within IFMTFM.............................146 7-2. Segment classification for M/D re action constants in ‘old’ IFMTFM...................156 7-3. Segment classification for M/D re action constants in ‘new’ IFMTFM.................156 7-4. Error associated with sample MeHg data points....................................................161 7-5. Parameters unchanged between ‘o ld’ and ‘new’ IFMTFM versions.....................161 7-6. Comparison of channel bank MeHg modeling in ‘old’ and ‘new’ IFMTFM........161 8-1. Comparison of steady-state velocity and depth output from SMS and RIVMOD................................................................................................................171

PAGE 9

ix LIST OF FIGURES Figure page 1-1. Map of the Carson River drainage and selected sampling sites.................................3 1-2. General locations of known mill taili ngs deposits and Scout Camp field site...........7 2-1. The aquatic mercury cycle.......................................................................................11 2-2. Processes that transport mercury into the water column along the Carson River....18 3-1. Modified channel geometry.....................................................................................32 4-1. Water table response to changing river stage...........................................................50 4-2. State variables measured with MERC4....................................................................54 4-3. Dependency of methylmercury concentration on soil moisture...............................57 5-1. Spatial location of survey points an d cross-sections at Scout Camp site.................65 5-2. Original survey points after import into SMS scatter module..................................72 5-3. Bathymetry of Scout Camp reach after mesh editing found in Appendix A...........75 5-4. Minimum flow levels able to be mode led, subject to a 0.2-meter head gradient.....76 6-1. 2001 USGS streamflow gage comparison: 15-minute data.....................................81 6-2. 2002 USGS streamflow gage comparison: 15-minute data.....................................81 6-3. May 2002 Carson River str eamflow at USGS gages...............................................82 6-4. Complete water table data from Well 1 and Well 3 at Scout Camp site..................87 6-5. Transient water levels in Well 1 and Well 3, 2001-2002.........................................89 6-6. Hydraulic differential from Well 1 to Well 3, 2001-2002.......................................90 6-7. Pump test drawdown data collected for Well 1 in October 2002.............................91

PAGE 10

x 6-8. 2001 streamflow routing comparison: FC H gage data and RIVMOD simulation for upstream boundary condition.............................................................................96 6-9. 2002 streamflow routing comparison: FC H gage data and RIVMOD simulation for upstream boundary condition.............................................................................97 6-10. April 2002 streamflow routing comp arison: FCH gage data and RIVMOD simulation for upstream boundary condition...........................................................97 6-11. 2001: near-channel well respons e to streamflow variability....................................99 6-12. 2002: near-channel response to streamflow variability..........................................100 6-13. April 2002: near-channel respon se to streamflow variability................................101 6-14. April 1-7, 2002: near-channel res ponse to streamflow variability.........................101 6-15. April 8-14, 2002: near-channel res ponse to streamfl ow variability.......................102 6-16. April 15-17, 2002: near-channel res ponse to streamflow variability.....................102 6-17. May 2002: near-channel response to streamflow variability..................................103 6-18. May 1-7, 2002: near-channel res ponse to streamflow variability..........................103 6-19. May 8-14, 2002: near-channel res ponse to streamflow variability........................104 6-20. May 15-21, 2002: near-channel res ponse to streamfl ow variability......................104 6-21. May 25-31, 2002: near-channel res ponse to streamflow variabililty.....................105 6-22. February 2002: near-channel re sponse to streamflow variability..........................105 6-23. February 1-3, 2002: near-channel re sponse to streamflow variability...................106 6-24. Hydrodynamic routing through RIVMOD se gments within reach of interest.......108 6-25. Water surface elevation (WSE) as a function of flow for March 2001 data..........109 6-26. March 9, 2002: location of surveyed water surface observations in RIVMOD segment 152............................................................................................................111 6-27. March 9, 2001: graphical approximati on of channel water surface slope in RIVMOD segment 152..........................................................................................111 6-28. Zero time lag: temporal correlation betwee n river stage and water table response in Well 1..................................................................................................114

PAGE 11

xi 6-29. Zero time lag, flow spikes removed: tempor al correlation between river stage and water table response in Well 1.........................................................................116 6-30. One hour time lag: temporal correlation between river stage and water table response in Well 1..................................................................................................117 6-31. Two hour time lag: temporal correlati on between river stage and water table response in Well 1..................................................................................................117 6-32. Channel thalweg location based on su rvey points from March and July, 2001.....121 6-33. Linear regression of March, July 2001 surveyed bed elevations for use in determination of hydraulic gradient for SMS model.............................................121 6-34. Stage – Discharge Relationship for St eady State Lowering of SMS Water Surface Elevation...................................................................................................125 6-35. Linear regression for determination of discharge as a function of depth...............127 6-36. Linear regression for determination of depth as a function of discharge...............129 6-37. 2001 rating curve-based estimates for water surface elevation in the channel compared to water ta ble level in Well 1.................................................................131 6-38. 2002 rating curve-based estimates for water surface elevation in the channel compared to water ta ble level in Well 1.................................................................131 7-1. Station and segment de finition within IFMTFM...................................................140 7-2. Conceptual IFMTFM segment interaction.............................................................142 7-3. Relationship between temporally-avera ged RIVMOD stage and bank saturation height.....................................................................................................150 7-4. Medium flow (May 16, 1994) IFMTFM calibration for inorganic Hg(II) (parameter 1) in channel banks for ‘old ’ and ‘new’ code versions......................153 7-5. High flow (June 10, 1995) IFMTFM verifi cation for inorganic Hg(II) (parameter 1) in channel banks for ‘old ’ and ‘new’ code versions......................154 7-6. Low flow (June 16, 1994) IFMTFM veri fication for inorganic Hg(II) (parameter 1) in channel banks for ‘old ’ and ‘new’ code versions......................154 7-7. Medium flow (May 16, 1994) IFMTFM calibration for MeHg ( 2) in channel banks (‘old’ code version) and K20 methylation rate in channel banks (‘new’ code version)..........................................................................................................158

PAGE 12

xii 7-8. High flow (June 10, 1995) IFMTFM calibrati on for MeHg threshold height (HT) in channel banks (‘old code version) and verification for MeHg in channel banks (‘new’ code version).......................................................................158 7-9. Low flow (June 16, 1994) IFMTFM verificati on for MeHg in channel banks for ‘old’ and ‘new’ code versions..........................................................................159 7-10. IFMTFM simulation for May / June 1995 flood event at Fort Churchill (model segment 140), illustrating MeHg dilu tion effect of threshold height HT in ‘old’ code version............................................................................................................163 7-11. IFMTFM simulation for May / June 1994 flows at Fort Churchill (model segment 140), demonstrating time lag for Me Hg response in ‘new’ code version....................................................................................................................163 8-1. Comparison of channel geometry at well transect for SMS and RIVMOD...........170 8-2. Q = 109.6 cms: comparison of steadystate velocity output from SMS and RIVMOD................................................................................................................173 8-3. Q = 6.85 cms: comparison of steadystate velocity output from SMS and RIVMOD................................................................................................................173 8-4. Q = 109.6 cms: comparison of steady-st ate depth predictions from SMS and RIVMOD................................................................................................................175 8-5. Q = 6.85 cms: comparison of steadystate depth predictions from SMS and RIVMOD................................................................................................................175 8-6. Comparison of contacted bank height predictions from SMS and RIVMOD.......176

PAGE 13

xiii Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Engineering SIMULATION OF SPATIAL AND TEMPORAL VARIABILITY OF METHYLMERCURY CONCENTRATION WITHIN CHANNEL BANKS AND SURFACE WATERS OF THE CARSON RIVER, NEVADA By Daniel E. Crawford August 2003 Chair: John J. Warwick Major Department: Environmental Engineering Sciences The USEPA designated the Carson River, Ne vada as part of a Superfund site in 1990. Between 1860 and 1890, elemental mercury was used at over 70 mill sites to recover gold and silver via mercury amalga mation, introducing 7000 tons of mercury to the Carson River watershed. Cont aminated sediments line channel banks downstream of these mill sites, with mercury concentrations two to three orders of magnitude above regional background levels. Mercury concen trations in both the Carson River and Lahontan Reservoir are among the highest known worldwide. Mercury methylation within sediments (accompanied by diffusi on, advection, and erosion processes) introduces toxic methylmercury to the water co lumn for ultimate transport into Lahontan Reservoir. Methylmercury concentrations in the bank are not spatially and temporally constant; but are instead a f unction of moisture history. A computer simulation model, Integrated Fluvial Mercury Transport and Fate Model (IFMTFM), was developed to simula te bank erosion and diffusion from bottom

PAGE 14

xiv sediments. A new bank element was develope d and incorporated into the previous IFMTFM, to relate flow history within the channel to moisture-content levels observed within the channel banks. Zones of va rying moisture affect methylation and demethylation kinetics; and therefore affect temporal concentrations of methylmercury available to the water column. Global positioning system (GPS) data were used to construct a detailed, temporal model of three-dimensional river hydrodynamics for a specific reach of the model domain, using Surface Water Modeling System (SMS) software. Streamflow and water table data show a tight coupling between the Carson River and the surficial aquifer. The nearly instantaneous water table response to streamflow oscillations (as observed in a monitoring well set back 5 feet from the ch annel bank) serves as the foundation for bank element functionality. Incorporation of the bank element into the complete Carson River model of hydrodynamics, sediment transport, and me rcury transport and cycling (IFMTFM) provides a new capability to simulate tempora lly variable mercury concentrations within the channel bank; reduces the number of mode l calibration parameters from three to two; and increases the accuracy of in-stream mercury predictions. The height of bank saturation and zone of active microbial methylation are co ntinuously updated as river stage changes. Contacted bank sediments re lease mercury to the water column through erosion, advection, and diffusion processes; with no threshol d height for methylation. Hydrodynamic components of IFMTFM (RIVM OD) and SMS were compared to show the value of high-reso lution transect data.

PAGE 15

1 CHAPTER 1 RESEARCH LOCATION: AN INTRODUC TION TO THE CARSON RIVER Carson River Basin The Carson River, located in the southwes tern corner of Nevada, drains from the eastern Sierra Nevada Mountains just sout h of the Lake Tahoe Basin (NDWP, 1996). The Sierra Nevada Mountains thus serve as the river basin’s prim ary source of precipitation. Precipitation in the Carson River Basin is attenuated by the rainsh adow of the Sierra Nevada. In the modeled portion of the Carson River (downstream of Carson City, Figure 1-1), the climate is semi-arid, with typical annual precipitation of less than 5 inches (Covay et al., 1996). Vegetation near the river is largely sagebrush or similar species, with cottonwood trees growing along the riparian zone immediately adjacent to the river. The Carson River Basin encompasses an area of approximately 3,966 square miles in the states of California and Nevada (Fi gure 1-1). The basin is topographically closed. A closed basin has no outlet, and water en tering such a basin discharges only by evapotranspiration or loss to groundwater (Covay et al., 1996). The upper portion of the Carson River Basin is drained by the Carson Ri ver’s East and West Forks. The Carson River's two forks merge in the northern part of Carson Valley and form the Carson River mainstem, which then continues on toward th e river system's terminus in the Carson Sink. The Carson Valley agricu ltural area diverts water from both river forks and from the Carson River mainstem, to provide irrigation for approximately 35,000 acres. Lahontan Reservoir, completed in 1915 with th e construction of Lahontan Dam, serves as a major sediment trap upstream of the Ca rson Sink. The Truckee Canal diverts water

PAGE 16

2 from the Truckee River Basin in to the reservoir, to meet the agricultural water demands associated with the Newlands Irrigation Pr oject (downstream of Lahontan Dam). United States Geological Survey (USGS) streamflow gages are available at Carson City (CCG station #10311000), Fort Churchill (FCH #10312000), Lahontan Dam (LDM #10312150), and the Truckee Canal (TRC #10351400). Of the Carson River Basin's total surf ace area, approximately 606 square miles (387,840 acres; or just over 15%) lie within the State of California. The remaining 3,360 square miles (2,149,680 acres; or almost 85%) li e within the State of Nevada (NDWP, 1996). While most of the basin’s surface area is in the State of Nevada, most of the basin’s precipitation originates in the Stat e of California. The geographic separation between the Carson River Basin’s principal s upply of water and its principal demands has historically resulted in cont roversy surrounding the rights to and uses of water resources. The lower Carson River Basin begins at th e head of Dayton Valley and ends at the furthermost extent of the Carson Sink (NDWP, 1996). The geology of this region (Lahontan Valley), the bottom of a prehistoric lake, is characterized by sediment deposits between 3,000 and 8,000 feet in thickness. Th e present-day Lahontan Valley and Carson Desert, now considerably drie r, is characterized by a num ber of important and varied hydrologic features including lakes, mars hes and wetlands, cultivated farmlands and pastures, barren desert sinks alkali flats, and playas. Carson River Streamflow The rain-shadow effect of th e Sierra Nevada Mountains to the west allows little precipitation to fall within the Lahontan Valley (NDWP, 1996). Average annual precipitation is only about 5 inches, while annual potential evaporation exceeds 60 inches.

PAGE 17

3 Figure 1-1. Map of the Carson River drainage and selected sampling sites. (Reprinted with permission from Carroll, R.W.H., 1999. Simulation of Mercury Transport and Fate in the Carson River, Nevada. Masters Thesis. University of Nevada-Reno, Reno, Nevada. Figure 1.1, pp. 9)

PAGE 18

4 Before extensive upstream storage and wa ter diversions, the Lahontan Valley was generally inundated in late winter and early spring by th e unhindered runoff of the melting snow pack from the Sierra Nevada Range. After this runoff period, a drying pattern would ensue during the remainder of the year until the wint er precipitation period commenced again in October-November. This highly seasonal pattern of wet and dry has been disrupted since the completion of the Newlands Project in the valley in 1903, and particularly after the completion of Lahontan Reservoir in 1915. Carson River flow is primarily from snow melt, but the effects of upstream storage and storage in Lahontan Reser voir have tended to moderate flow magnitudes. The cycle of seasonal high flows in the spring (due to snowmelt) and low flows during the remainder of the year in the Carson River ups tream of the Lahontan Reservoir continues, but the flow magnitudes no longer reflect pre-development conditions. Snowmelt, whether immediate or from upstream reservoi r releases, is the source of Carson River streamflow; this observation, along with the se mi-arid Nevada climate and the extensive diversion of water for agricu ltural use along the entire Carson River, results in a highly variable streamflow pattern. Carson River flows may vary by three to four orders of magnitude on an annual basis. During late summer and early fall, streamflow in downstream reaches of the Carson River may be nonexistent. By contrast consider the rain-on-snow flood event of January 1997; the Carson River extended well into the floodplain, with peak discharge values of 24,000 cubic feet per second (cfs ) and 22,300 cfs estimated at Carson City Gage (CCG) and Fort Churchill Gage (FCH), respectively (USGS, 1997).

PAGE 19

5 Carson River Aquifer Systems Principal aquifers in the Carson River Ba sin are unconsolidated basin-fill deposits of interbedded gravel, sand, silt, and clay (C ovay et al., 1996). Aquifers within the basin readily receive, transmit, a nd store large volumes of water. Detailed hydrologic investigations of the Carson River Basin ha ve been performed as part of the USGS National Water-Quality Asse ssment (NAWQA) program. General areas along the Carson River underlain by basin-fill aquifers in clude Carson Valley, Eagle Valley, Dayton Valley, and Churchill Valley. Unconsolidated deposits occur in all valle ys of the Carson River Basin and are the primary water-bearing units (Covay et al., 1996 ). Carson Valley contains unconsolidated Quaternary basin-fill sediment s (mostly fineto coarse-grained Carson River alluvium), and semiconsolidated Tertiary sediments (m ostly lake and stream deposits) (Maurer, 1986). Eagle Valley contains Quaternary semi consolidated and unconsolidated alluvium (including deposits of sand, silt, and coarse materials; and cl ay confining units) (Trexler, 1977; Arteaga, 1986). Basin-fill deposits in Dayton and Churchill Valleys include Cenozoic sedimentary deposits ( 500 to 3000 ft thick); alluvial fans and pediments; and lake, river, delta, and floodplain deposits (consisting of a mixtur e of clays, silts, sands, and gravel) (Schaefer and Whitney, 1992). The sediments within the modeled reach of the Carson River Valley clearly may be defined as typical alluvium. Historical Mining Operations Historical mining activities have contamin ated lower reaches of the Carson River, Lahontan Reservoir, and Stillwater Mars h with mercury. Extensive milling and ore processing operations were situated within Dayton Valley during the era of the Comstock Lode during the late 1800s (NDWP, 1996). Here the Carson River passes through a

PAGE 20

6 canyon and loses 240 feet of elevation over a di stance of 10 miles. With its proximity to the Comstock mines to the north and the av ailable hydropower, this was an ideal location to establish the mills. During the peak of the Comstock era, more than 75 mills operated along the Carson River. Gold Canyon and Sixmile Canyon (Figure 1-2) mark the geologic boundaries of one of this nation's richest silver strike s—the Comstock Lode. Known mill tailings deposits are found in areas along the Carson Ri ver from just downstream of Carson City to Dayton and in areas along the Carson River tr ibutaries from Virginia City (Wayne et al., 1996), as shown by the crosshatched areas in Figure 1-2. “Efforts to extract this vast body of silver wealth would begin a period of natural resource exploitation (timber cutting and water diversions) and environmen tal degradation (defor estation, soil erosion, water pollution, arsenic and merc ury spills), the eff ects of which have carried forward to the present day” (NDWP, 1996). During the Co mstock's peak years of production (1862 to 1880), approximately $296,400,000 worth of silv er and gold were removed from the mines. Over its entire active mining period, which lasted from 1857 to 1921, the Comstock yielded $386,346,931 (historical value) in total mineral production (of which $222,315,814 was in silver; $164,023,917 was gold; and $7,200 was copper and lead). Mining operations associated w ith the Comstock Lode resulted in the production of an extensive volume of mine tailings (waste ro ck material). It has been estimated that during some peak periods of mining, more th an 600 tons of tailings material were released to the surrounding ar ea on a daily basis (Miller et al., 1998). Liquid elemental mercury was used to extract gold and silv er from the mined material using an amalgamation process. This introduced an estimated 7000 tons of mercury into the

PAGE 21

7 Figure 1-2. General locations of known mill tailings deposits and Scout Camp field site Carson River system through mercury residual within the mine tailings (Carroll, 1999), along with the immense, immeasurable volum e of associated sediment. At the mills, mined ores were pulverized and mixed with mercury, forming a mercury-silver/gold amalgam because of the high affinity of merc ury for these metals (a method known as the Washoe process). Float tanks were used to separate the amalgam from the waste rock. Tailings deposits Scout Camp field site

PAGE 22

8 Roasting the amalgam, releasing the mercury from the complex as a vapor, would then produce a pure form of the precious meta ls. Although most of the mercury vapor was recovered, it was estimated that as much as 0.68 kg of mercury was released to the environment with mill tailings for every ton of ore that was processed (Miller et al., 1996). During the past 140 years, mercury-cont aminated sediments were eroded from the mill sites, transported downstream by fluvial processes, and re-deposited in specific localities, thereby expanding the influen ce of mercury pollution (Miller et al., 1996). Carson River Geomorphology Analysis of historical documents and a lluvial stratigraphy indicate that the Carson River underwent significant changes in channel form and process during the period of the Comstock Lode (Miller et al ., 1998). The increased sediment load resulting from the constant supply of mine tailings, along with sediment introduced from clear-cut headwater areas used for timber production during this period, caused the river to aggrade (on the order of several meters); and to experience lateral instability and meander abandonment. The valley floor and abandone d channels subsequently filled with contaminated debris. The end result of these pro cesses was the storage of mercury-enriched sediments within a complexl y structured alluvial sequence located all along the Carson River valley downstream of Ca rson City (Miller et al., 1999). Mercury concentrations in the alluvial valley fill deposits exhibit great spatial variability (longitudinally and laterally along the river; and vertically wi thin sediment deposits), but the concentrations are generall y two to three orders of magn itude greater than regional background levels of 0.01 to 0.10 g/L (Miller et al., 1998).

PAGE 23

9 The eventual reduction of mining operati ons decreased the sediment load and caused the river to downcut through these c ontaminated sediments. Today, the Carson River continues to reclaim its original fl oodplain in an attempt to establish a new equilibrium. Mercury-contaminated sediment is believed to enter the system primarily during higher flow events when water leve ls reach overlying contaminated sediments (notably within the floodplain) and erosi onal processes become significant. Initially, mercury released from mines represented a point-source contaminant problem. Modern mercury pollution of the Carson River must be considered a dist ributed-source problem because of the contaminated sediment stored within channel banks.

PAGE 24

10 CHAPTER 2 MERCURY CONTAMINATION AND MERCURY CYCLING Mercury Cycling within Surface Water Systems Mercury is introduced to surface waters through atmospheric deposition, surface runoff (erosion), and diffusion. After mercury enters into surface water, it enters a complex cycle of transformations—primarily according to redox chemistry and microbes. A schematic drawing of mercury cycling in an aquatic ecosystem is shown in Figure 2-1 (USGS, 1995). Mercury may remain in the wate r column or be lost through drainage or advection. Mercury can be incorporated in to the bed sediments by particle settling, perhaps only to be later released by diffu sion or resuspension. The formation of methylmercury can introduce mercury into th e food chain, or volatili zation may exile the mercury back to the atmosphere. Divalent mercury and methylmercury, critical components of the Carson River system, can be transported to water bodies in runoff, typically bound to suspended soil/humus or a ttached to DOC (dissolved organic carbon), or Hg(II) and methylmercury can leach into the water body from groundwater flow in the upper soil layers (USEPA, 1997). Research has shown that both DOC and pH can play a significant role in the fate of mercury with in a system. Studies have shown that DOC concentration and pH levels have a strong e ffect on mercury fate, w ith increasing acidity (decreasing pH) and increasing DOC content resulting in higher body burdens in fish (USGS, 1995). The movement of mercury will vary within each given body of water. Mercury concentrations are typically reporte d for the water column (dissolved or

PAGE 25

11 particulate), the underlying sediment (surface or deep sediments), and biota (generally fish) (USEPA, 1997). Quantification of Mercury Levels in the Carson River System The major water issue for the lower Carson River Basin focuses not on quantity but rather quality (NDWP, 1996; Carroll, 1999; Miller et al., 1998). Me rcury is introduced into the river from the contaminated mine tailings distributed throughout the riverbank sediments and floodplain deposits. Mercury concentr ations in the tailin gs are two to three orders of magnitude greater than background concentrations in the region. Sampling by the United States Environmental Prot ection Agency (USEPA) found mercury concentrations as high as 4,600 parts per milli on (ppm) in the fine-grained materials of the tailings piles. Mercury dispersal from th ese tailings is extensive. Concentrations exceeding 100 ppm have been measured more than 60 kilometers (km) downstream of Figure 2-1. The aquatic mercury cycle. (USGS, 1995. Mercury Contamination of Aquatic Ecosystems. U.S. Geological Surv ey Fact Sheet 216-95. Figure 6, pp. 3)

PAGE 26

12 the mining operations. During the most inte nse period of mining activity, the Carson River emptied directly onto an extensive wetland now known as the Stillwater Wildlife Management Area (SWMA). The completion of Lahontan Dam in 1915 created a new settling basin for sediments derived from the mercury-contaminated mine tailings, but a large amount of mercury had already been in troduced to the wetla nd habitat downstream of the dam. Several extensive site-characterization studies addressed mercury contamination along the Carson River System. Total mercury (Hg) in three mountain streams indicates background levels of < 2 nanograms per liter (ng/ L) for this region of Nevada (Wayne et al., 1996). Water samples collected in 1992 and 1993 showed the following: Elevated (e.g., > 20 ng/L) mercury le vels first appear downstream from accumulations of mill tailings. Total mercury concentrations in unfilter ed and filtered water increase downstream. Mercury concentrations in both the Carson River and Lahontan Reservoir are among the highest known worldwide (100 to 1000 ng/L); filtered water samples suggest that the >0.4 m particle fraction constitutes over 60 % of the total water-borne mercury (Wayne et al., 1996). Abnormally high levels of Hg, Au, and Ag occur in sediments downstream of former ore milling operations. Systematic trends are apparent from the analysis of bank and bed samples (Wayne et al., 1996): mine tailings have the hi ghest Hg, Au, and Ag contents, and fine-grained silt s and muds from the river ba nks typically have higher Hg contents than coarser sands and gravel from the river channel; and mercury contents in both channel and bank deposits increase with increasing distance from CCG. Potential explanations for the high mercury levels in the Carson River system when compared to other contaminated sites include the large quan tity of elemental mercury introduced to the

PAGE 27

13 system and the time period of over 100 years allowed for physical dispersal and chemical alteration (Wayne et al., 1996). Introduction of Mercury to the Carson River System The original elemental form of mercury released to the environment remains the prevalent form in highly contaminated soils, sediments, and taili ngs (Lechler et al., 1997). Subaerial (buried) tailings typically c ontain greater than 95% elemental mercury; downstream of the influence from the tailings along the Carson River, the percentage of total mercury as elemental mercury increases to slightly greater than 90% before decreasing further downstream (Lechler et al., 1997). Elemental mercury possesses a high volatility and low water solubi lity, and equilibrium conditions of exposed sediments does result in both volatilization and some dissoluti on. The species of primary interest in terms of mercury toxicity are the organo-merc ury compounds such as methylmercury (CH3Hg+ or MeHg) (Wayne et al., 1996). However, methylmercury is the end product of the methylation of inorganic diva lent mercury (Hg(II) or Hg2+). Elemental mercury (Hg(0) or Hg0) is the mercury form introduced to the Carson River System. The low reactivity of the prevalent Hg(0) species does not allow for direct methylation of this species. Therefore, the environmental transformations of elemental mercury are next considered. Starting with elemental mercury, the two potential pathways for mercury oxidation involve volatilization and dissolution. Following its introduction a nd transport in the atmosphere (after volatilization), Hg(0) (g as) can undergo transformation to Hg(II) (gas) and Hg(II) (particulate) within cloud wate r (USEPA, 1997). This mechanism has been proposed as an explanation for the global natu re of mercury pollution, as it can result in mercury deposition to land and water. The ci rculation of air masse s on the global scale has the potential to take Hg(0) emissions from their point of origin and carry them to

PAGE 28

14 anywhere on the globe before transformation and deposition occur. The slow oxidation of Hg(0) to Hg(II) results in Hg(0) having an at mospheric residence time on the order of 1 year. Other forms of mercury ar e likely deposited before they are ever able to dilute within the global atmosphere. The end result of these atmospheric transformations is that mercury released to the atmosphere deposits mainly as Hg(II), whether from the direct deposition of emitted Hg(II) or transformati on of emitted Hg(0) to Hg(II) by the above process (USEPA, 1997). Elevated depositi on rates therefore occur around emission sources, but regional or global tr ansport and deposition also o ccurs. In other words, even though the principal emissions of mercury ar e from point sources concentrated in industrial areas, mercury polluti on is truly a global concern. While this is a pathway that does occur, the arid climate, and hence lack of rainfall, of the Carson River region and the likelihood of wider atmospheric transport make it an unlikely source for the large-scale introduction of Hg(II) to this system. Rather, a small fraction of the liquid Hg(0) that is associated with the mill tailings is dissolved out of the amalgam particles and subsequently adsorbed to fine-grained sediments. These sediments are then deposited in downstream, low-energy reaches of the river and reservoir (Lechler et al., 1997). Research by de Magalhaes a nd Tubino (1995) has shown that Hg(0) can be easily oxidized by molecular oxygen in aqueous so lution in the presence of species such as chloride, which complex Hg(II); high chlo ride activities in Ca rson River and Lahontan Reservoir sediments are a product of the prevailing arid climate conditions and subsequent evapo-concentration processe s (Wayne et al., 1996). This mechanism facilitates mercury oxidation to occur in sediments that remain moist or in suspension, including channel bed sediments and bank se diments under certain moisture conditions.

PAGE 29

15 Other oxidizing agents within the sediment matr ix and in the water, however, are likely at work as well. The presence of sulfur within the ores of the Comstock era results in the precipitation of some of the oxidized divalent mercury as relatively insoluble HgS in highly-reducing environments, such as within the reservoir or buried portions of the soil profile (Lechler et al. 1997). This mechanism serves to limit methylation of all oxidized mercury, although very limited methylation of HgS may occur. Note that the prevalent mercury species within the system remains the elemental form. The key item to take from the above discussion, though, is the presence of inorganic, divalent mercury (Hg(II) or Hg2+) within the near-channel a nd in-channel sediments of the Carson River system. Chemistry of Mercury in Soil Divalent mercury within soil is subject to many chemical and microbial reactions (USEPA, 1997). For typical ranges of pH, te mperature, and humic content in soils, conditions generally favor the formation of inorganic Hg(II) compounds such as HgCl2, Hg(OH)2, and inorganic Hg(II) compounds complexed with organic anions. These inorganic Hg(II) compounds possess high water so lubility, but this theo retical mobility is heavily limited by the tendency of these com pounds to form complexes with soil organic matter, primarily humic and fulvic acids, because of the affinity of Hg(II) and its inorganic compounds for sulfur-containing functional groups. Because of the high affinity of mercury species for soil, soil acts as a large rese rvoir for mercury introduced to the environment and a leaching source that will continue for many years after the cessation of mercury loading. Partition coefficients (Kd = concentration of Hg in dry sediment, soil, or suspended matter divided by the dissolved concentrat ion in water) have been calculated for the relative affinity of Hg(II) and MeHg for sediment over water. Typical values are 10 to 100,000 milliliters (mL) per gram(g) of soil, 100,000 mL/g of

PAGE 30

16 sediment, and 100,000+ mL/g for suspended matte r, indicative of a strong preference Hg(II) and MeHg to remain in these phases (USEPA, 1997). In aquatic systems with significant mercury loadings, it is possible fo r a significant amount of Hg(II) to partition into the water column, especially in those in stances where the concentration of suspended material in the water column is high, as in the case of a flood event. Alluvial sediments within the Carson River system are generally mi neral-rich soils with low organic content. In the absence of large amounts of organic material, complexation with these compounds is limited; studies have suggested that e nhanced mercury mobili ty within the Carson River and Lahontan Reservoir system may also re sult from locally high chloride activities (Wayne et al., 1996). Divalent mercury com pounds and complexes can undergo reduction to elemental mercury, processes mediated by both humic substances and light (USEPA, 1997). These processes will bring about diffusion of the gas and re-emission to the atmosphere. Microbial activit y on certain Hg(II) substan ces can result in MeHg formation, which typically makes up between 1 and 3 percent of mercury found in soil (USEPA, 1997). Most of the remaining merc ury (97 to 99%) will be present as Hg(II) species. Contaminated sediments may contain Me Hg levels far greater than these typical observations. Mercury Transformation and Toxicity Heavy metals, including mercury, possess a strong affinity for sorption onto particulate matter given the Eh and pH conditions found in most aquatic environments (Morel and Hering, 1993). Several contamin ated river systems convey roughly 90% of their total heavy metal load via sediment tr ansport (Babiarz at al., 1998; Balogh et al., 1997; Gill and Bruland, 1990; Hurl ey et al., 1995; Meade et al., 1995). The Carson River

PAGE 31

17 is no exception in that more than 95% of total mercury can be removed from water column samples by filtration (Bonzongo et al., 1996a). Several forms of mercury are present within the sediments and waters of the Carson River system. Due to its high volatility, only very low concentrations of elemental mercury are detected in the surface waters of the Carson River (Bonzongo et al., 1996a). Inorganic mercury and methylmercury in sediments along the Carson River are transported into the water column through di ffusion and erosion processes (Figure 2-2). Mercury bound to particulate ma tter is generally considered unavailable for any chemical or biological reaction and will be susceptible to sedimentation processes. The presence of a state of sorption equilibrium (partitioning between sorbed and dissolved phases) does, however, ensure the presence of a sma ll quantity of free ionic mercury (Hg2+) within the water column and pore water of saturated bed sediments and moist bank sediments. Free ionic mercury is available for transforma tion. Within bottom sediments, a positive but weak correlation between the ratio of total Me Hg to total Hg and organic matter has been observed (Hoffman and Thomas, 2000). Sulfate-reducing bacteria (SRB) methylat e free ionic mercury into the organic mercury species methylmercury (CH3Hg) and dimethylmercury ((CH3)2Hg) (Gilmour and Henry, 1991; Gilmour et al., 1992; Ma tilainen, 1995; Bonzongo et al., 1996a). The production of MeHg by SRB is primarily controlled by SO4 2concentrations required for SRB activity. Optimal concentrations for the biotransformation of inorganic mercury by SRB are estimated in the 200 to 500 M/L range; above this range, methylation is inhibited by high production of sulfide, while lower concentrations inhibit methylation because of sulfate availabili ty (Bonzongo et al., 1996a). Eviden ce is available to support

PAGE 32

18 Figure 2-2. Processes that tr ansport mercury into the wa ter column along the Carson River. (Reprinted with permission fr om Carroll, R.W.H., 1999. Simulation of Mercury Transport and Fate in the Ca rson River, Nevada. Masters Thesis. University of Nevada-Reno, Reno, Nevada. Figure 1.2, pp. 11) abiotic contributions to the methylation a nd demethylation processes as well, but the exact mechanisms remain poorly understood (Chen et al., 1996). While all mercury species possess intrinsic toxic properties, it is only the organic mercury species that are lipophilic and can bioaccumulate through the food chain (Fitz gerald and Clarkson, 1991; Zillioux et al., 1993). Within freshwater systems, methylmercury accounts for the majority, if not all, of the organic merc ury species; the lack of dimethylmercury accumulation in freshwater systems may be related to both its high volatility and photochemical destruction (Fitzgerald and Ma son, 1997). Methylmercury is therefore the principal species of concern for both ecosystems and human health. Sulfate-reducing bacteria are widespread in anaerobic bottom sediments but may also reside within the anoxic hypolimnion of large lakes (Gilmour and Henry, 1991; Matilainen, 1995; Watras et al., 1995b). Fluvi al systems contain mostly oxic waters; within the context of this study, only the a noxic channel bottom sediments and the moist

PAGE 33

19 or saturated bank sediments (potentially anoxic environments) will be considered as sites for mercury methylation. The reader should further note that methyl ation and demethylation processes occur simultaneously. Methylation refers to the biologicallyor abiotically-mediated transformation of free ionic (inorganic) Hg(I I) to MeHg; demethyla tion is the conversion of MeHg back to inorganic Hg. While demethylation reactions are kinetically unfavorable, due in part to the presence of partially covalent metal-to-carbon bonds, the unfavorable kinetics can be overcome th rough microbial enzymatic or photochemical mechanisms (Amyot, 1998). While both processes must be modeled, inte rest is centered on the net rate of MeHg production, as this is the determining factor for methylmercury concentrations in channel bottom and bank sediments and (through subsequent diffusion and erosion) the water column, the latter of wh ich is available for introduction to the food chain. The balance between methylation and de methylation processes may be critical to maintaining low levels of MeHg in some surface waters. Processes which reduce the availability of inorganic Hg(II) for methyl ation can further limit the potential for biological uptake. Examples of these proce sses include the followi ng: sedimentation of inorganic Hg(II) compounds resulting in th e precipitation of insoluble HgS as the sediments become increasingly anaerobic; photochemical or microbial reduction of Hg(II) to Hg(0), which is removed by volat ilization, within aerobi c surface waters; and microbial or abiotic reduction of Hg(II) within anaerobic sediments (Amyot, 1998). Methylmercury in the Water Column The result of the chemical transformation of free ionic Hg2+ to MeHg is the introduction of methylmercury into the water column by diffusion and erosion processes. Large flow events, when the Carson River er odes significant quanti ties of contaminated

PAGE 34

20 bank material, are believed to be the princi pal pathway by which mercury enters into the water column (and ultimatel y the Lahontan Reservoir a nd Wildlife Refuge areas). Because mercury is adsorbed to the sediment the increased erosion associated with high flow events and the higher quantity of me rcury present within the floodplain (compared to the channel bottom) results in more mercur y getting into the water column. Particulate Hg is typically greater than 50% of total Hg and increased downstream; by contrast, the dissolved fraction of MeHg constituted a larger fraction of total MeHg than particle-associated MeHg (Bonzongo et al., 199 6b). Flood events also contribute greatly to the continued dispersal of mercury contaminated sediments throughout the river system. Concentrations of both total Hg and MeHg in water are strongly influenced by flow regimes and well correlated with concentr ations of suspended solids, as result from erosion (Bonzongo et al., 1996a; Bonzongo et al., 1996b). Observations of MeHg and total Hg in the water column and bank sediments indicate that the MeHg and total Hg concen trations measured in surface waters were discharged in large part from Hg-contaminated bank materials, which are not in permanent contact with the water body (Bonz ongo et al., 1996a). Contaminants entering the water column that are associ ated with silt and clay are transported to the reservoir, while mercury associated with amalgam gr ains and other sand-sized sediment are incorporated into the materials found ben eath the channel floor (Miller et al., 1998). Mercury also enters the water column by di ffusion from the channel bed and contacted bank sediments (Carroll, 1999; Ca rroll et al., 2000; Wayne et al., 1996). This source term is thought to be particularly important for the introduction of MeHg to the water column when the discharge is low and erosion is minimal. For water samples, positive but weak

PAGE 35

21 correlations between total MeHg concentration and water temperature, pH, and total organic carbon have been obs erved; in addition to a pos itive, moderate correlation between total MeHg and total Hg concentr ations. A strong positive correlation between total MeHg in water and total Hg in sedime nt has also been observed (Hoffman and Thomas, 2000). The distribution and occurrence of Hg speci es in aquatic environments are probably regulated by key limnological and chemical pa rameters. While significant concentrations of methylmercury (MeHg) exist within the surface waters an d biota of the Carson River, the ratio of total MeHg to total Hg is much smaller than other contaminated sites (Bloom and Effler, 1990; Bonzongo et al., 1996a; Gill and Bruland, 1990). The ratio of total MeHg to total Hg decreases downstream, reaching values of less than 1% below the mine tailing sites. This suggests both a high percen tage of inorganic mercury input from point sources and a low net product of methylat ion-demethylation within the system (Bonzongo et al., 1996b). There is a strong indication that net methylation rates increase with decreasing pH in both the water column and at the sediment-water interface (Watras et al., 1995b; Xun et al., 1987). The alkaline nature of the Carson River (pH 7.5 to 9.0) may limit the MeHg production. Naturally occurring Group VI oxyanions (W, Mo, and Se) also inhibit the methylat ion process by SRB, and the extremely high levels of inorganic mercury found in the Carson system during winter and spring flood events may actually be toxic for SRB as well (Bonzongo et al., 1996a; Chen et al., 1996; Chen et al., 1997). The Carson River has a very low organic co ntent (<2%) (Miller et al., 1998; Wayne et al., 1996), and no correlation exists betw een DOC and total mercury or methylmercury

PAGE 36

22 concentrations (Bonzongo et al., 1996a). Lo w levels of DOC limit complexation of dissolved MeHg to dissolved organic ligands and lead to the assumption that a large fraction of dissolved MeHg is available fo r conversion processes and uptake by aquatic organisms (Bonzongo et al., 1996b). Increase d temperatures result in increased methylation rates along the Carson River (Carro ll, 1999; Carroll et al ., 2000); this direct relationship is strongly supported with severa l studies of other systems (Callister and Winfrey, 1986; Matilainen, 1995; Stordal a nd Gill, 1995; Watras et al., 1995a). Methylmercury within the water column is available for introduction to the aquatic food chain. MeHg is introduced to the wate r column through bank erosion and diffusion from contacted channel bank and bed sediment s. Analysis of water samples from the Carson River indicates that only a very sma ll fraction of mercury found in the water is available for conversion processes, including methylation, demethylation, and Hg(0) (or Hg0) formation (Bonzongo et al., 1996b). Total Hg and total dissolved mercury concentr ations in the contaminated part of the Carson River are 5 to 100 times higher than the following recently studied systems polluted with mercury: Davis Creek, Califor nia; Clear lake, Ca lifornia; Lake San Antonio, California; Pyramid Lake, Nevada; Onondage Lake, New York; and Clay Lake, Ontario (Bonzongo, 1996b). Mercury concentrat ions in the Carson River are likely among the highest in any aquatic systems, even those with known histories of Hg contamination. However, the reactive fraction is observed to be lower within the Carson system. This data is used to justify the ex clusion of methylationdemethylation processes within the surface waters of the Carson River model, limiting these reaction sites only to bed sediments and saturated or moist bank sediments.

PAGE 37

23 Methylation and demethylation kinetics ar e variable spatially within the Carson River system and vary with the time of y ear. Preliminary findings from the USGS study on methylmercury formation and degradation in sediments of the Carson River system (CRS) include the following (Marvi n-DiPasquale and Oremland, 1999): High rates of MeHg production in upstream river sediments and low rates in wetland sediments; Low rates of MeHg degradation in the upstr eam river sediments and high rates in the wetlands; Gross MeHg degradation strongl y impacts net MeHg production; Microbial sulfate reduction a nd resulting reduced-S concentr ations play an important role in mediating Hg-transformations; Oxidative demethylation is the dominant degradation pathway in the CRS Agricultural drains are a poten tially important source of me rcury to the larger wetland areas; Distinct spatial trends in microbial Hgdynamics reflect the immediate environmental conditions with respect to organic matter, sediment redox, pore-water anions, and overall anaerobic microbial activity. Note that most of these observations were discussed at greater length within earlier portions of this document. Mercury in the Aquatic Food Chain The quantity of mercury transformed to methylmercury (methylated) can vary greatly between water bodies because of several factors; some of these factors, other than mercury loading rate, include the following: environmental setting (climate, geology, land use, land cover), water chemistry, and we tland density, with the latter being the most important determinant of methylmercury pr oduction at the basin scale (USEPA, 2000). Methylation is a key step in the entran ce of mercury into the food chain. This methylation process occurs principally th rough biological processes in the aquatic sediment or water column, but also perhap s through abiotic pro cesses involving humic

PAGE 38

24 and fulvic acids in solution. Demethylation re actions, volatilization of elemental mercury and dimethylmercury, and bonding to particulat e matter and settling pr event all of the mercury loaded into a system from being bioavailable in the aquatic environment (USEPA, 2000). Microbial uptake of mercury is the fi rst step for both possible methylation processes and bioaccumulation considerations Methylation by sulfate-reducing bacteria or entry into the aquatic f ood chain via phytoplankton or bacteria first mandates the transport of mercury across the lipid memb rane that surrounds unicellular organisms (Amyot, 1998). The microbial uptake of inor ganic Hg(II) and methylmercury are affected by elements of water chemistry. High chloride concentrations and low pH conditions that favor the uncharged chloride complexes HgCl2 and CH3HgCl, mercury complexes which possess the highest lipid solubility, have been shown in field data to promote the accumulation of inorganic Hg and MeHg at the microbial level by phytoplankton and bacteria. Other non-polar mercury species, including dimethylmercury and elemental mercury, also diffuse rapidly through cell me mbranes, but they are not bioaccumulated. Bioaccumulation refers to organisms taking up contaminants more rapidly than their bodies can eliminate them, allowing the am ount in the body to increase over time. In order for high concentrati ons to occur in fish, mercur y must be efficiently taken up by microorganisms at the bottom of the f ood chain, retained by these organisms, and passed on to their predators (Amyot, 1998). A lthough trace metals are often accumulated at the bottom of the food chain, concentratio ns in biomass do not often increase up the food chain (a process called biomagnificati on). Biomagnification is the incremental increase in contaminant concentration at each level of the food chain; this phenomenon is

PAGE 39

25 the result of higher concentra tion of contaminant in the f ood sources for organisms higher on the food chain (USGS, 1995). Methylmercury is very bioavailable and accumulates in fish through the aquatic food web, with nearly 100 percent of mercury found in fish muscle tissue in the methylated stat e (USEPA, 2000). Elemental mercury and dimethylmercury are not bioaccumulated at the micr obe level because they are non-reactive and tend to diffuse out just as easi ly as they diffused into the cell membrane. Further explanations are needed as to why MeHg bioaccumulates and Hg(II) does not, despite both HgCl2 and CH3HgCl diffusing into cells at similar rates, reacting with components of the cell, and being efficien tly retained (Amyot, 1998). Research was conducted to explain this phenomenon on cope pods feeding on marine diatoms. What was shown is that Hg(II) is primarily bound to particulate cellular material of the diatom, while MeHg is associated with the soluble portion within the diatom cell. Because the particulate material is excreted by the c opepod rather than absorbed, only the MeHg associated with the soluble fraction is effi ciently incorporated into the copepod. Other field data has shown this difference in tran sfer efficiency as applicable to other microorganism predator relationships, a lthough perhaps counter-intuitive based on the relative abundance of Hg (II) compared to MeHg found in water bodies. This process continues up the food ch ain through zooplankton, and the lipid solubility of MeHg allows it to be retained in fat tissues of these organisms. Further explanation is, however, needed since once the trophic level of fish is reached, MeHg burden in muscle tissue is of greater significance that in fat tissue (Amyot, 1998). In fish, MeHg absorption occurs principally along the in testinal wall, whereas inorganic Hg(II) is adsorbed at the microvilli interface (locat ed on immune system lymphocytes that are

PAGE 40

26 carried within the bloodstream); this result s in a very low uptake rate of inorganic mercury in comparison with methylmercury. This cycle results in increases in the proportion of MeHg to total Hg from about 10% in the water column, to 15% in phytoplankton, to 30% in zooplankton, and to 95% in fish. In fact, bioconcentration factors (BCF = concentration of MeHg in biota divided by concentration of MeHg in water column) are on the order of 105 to 106 (USEPA, 2000). The food chain structure therefore plays a key role in mercury bi omagnification, because the accumulation of MeHg results mainly from the consumption of MeHg-containing food rather than uptake from the water itself. As the number of tr ophic levels within an aquatic system is increased, so too is the mercury concentra tion in the top predators of the food chain (Amyot, 1998). Large saltwater fi sh with long life cycles, mo st notably sharks, tuna, and swordfish, are likely to accumulate th e highest amounts of MeHg (USDHHS, 1997). Higher methylmercury contents are also typical ly found in fish species in more acidic lakes (lower pH) (USEPA, 2000). Studies have shown that DOC concentration and pH levels have a strong effect on mercury fate, with increasi ng acidity (decreasing pH) and increasing DOC content resulting in higher body burdens in fish, perhaps by enhancing the mobility of mercury in the environment and thus the likelihood of entering the food chain (USGS, 1995). Even though mercury loadings to some remote areas may be very low, the strong absorption by microbial bact eria and plankton found in floating material make the possibility of biomagnificati on increasingly likely if other conditions are suitable (USGS, 1995). Although mercury is a globally di spersed pollutant, problems do not arise everywhere. With the exception of grossly polluted areas, mercury is generally only a

PAGE 41

27 problem where the rate of natu ral formation of methylmercury from inorganic mercury is greater than the reverse r eaction (USGS, 1995). Specific environments known to favor the production of methylmercury include certain types of we tlands, dilute low-pH lakes in the northeast and north cen tral United States, parts of the Florida Everglades, newly flooded reservoirs, and coastal wetlands, partic ularly along the Gulf of Mexico, Atlantic Ocean, and San Francisco Bay (USGS, 1995). Because methylmercury is the mercury species of concern with rega rd to bioaccumulation and biomagnification within the food chain, quantification of MeHg contaminati on within the Carson River is of great importance. Carson River Superfund Site In August 1990, the stretch of the Carson River through Dayton Valley and extending downstream to Lahontan Reservoi r, to include Carson Lake, Stillwater National Wildlife Refuge (NWR), and Indian Lakes in Lahontan Valley, along with adjoining tributaries stretching nearly to Virginia City it self, was declared a Superfund site by the U.S. Environmental Protection Ag ency (USEPA). The re mediation strategy for these sites remains unresolved. Lahontan Valley's wetland system encompasses a large portion of the lower Carson River Ba sin and includes the Stillwater NWR, the Fallon NWR, and Stillwater Wildlife Management Area. This area serves as a key production, migration, and wintering area for up to one million waterfowl, shorebirds, and raptors (NDWP, 1996). Merc ury concentrations in soil samples are found to be some 200 times greater below the mill sites than a bove them, with the highest concentrations found just upstream from Lahontan Reser voir (the principal area for sediment deposition). Fish within the Carson River syst em have been found to have total mercury levels four times greater than the Fe deral Action Limit for consumption of 1 g/g (Miller

PAGE 42

28 et al., 1996; Carroll, 1999). Mercur y levels in fish (up to 9.5 g/g ) and waterfowl (>3.6 g/g) caught in and near the SWMA (Stillwat er Wildlife Management Area) have also been found to be well above the FDA limits for human consumption. The recreational value of Lahontan Reservoir, which include s commercial and recreational fishing, and the SWMA are severely impacted by the c ontinued input of me rcury. Abnormally high Hg concentrations in sediments and fauna from the SWMA support a scenario of Hg mobilization, transport, and methylation w ithin the Carson-Lahontan system (Wayne et al., 1996). Health Effects of Mercury Contamination Airborne and waterborne me rcury may undergo several bi ological transformations that can yield methylmercury. In this form, mercury efficiently accumulates in the aquatic food web. Nearly all of the mercury that accu mulates in fish in rivers, lakes, and reservoirs is present in its methylated form. The methylme rcury concentrations found in fish specimens may be several orders of ma gnitude greater than c oncentrations in the water column, the direct result of the bi oaccumulation and biomagnification processes. Inorganic mercury tends to not bioacc umulate (USEPA, 2000). The efficient accumulation of methylmercury up the aquatic food web poses risks to both humans and ecosystems. Other forms of mercury can also impact human health. Toxic effects associated with methylme rcury include permanent damage to the brain and kidneys; and damage to the ner vous system (USEPA, 2000). Mercury effects many different brain functions and symptoms may occur wh ich include the following: personality changes (irritability, shyness, ne rvousness), tremors, changes in vision or hearing, loss of sensation, and memory loss (United States Department of Health and

PAGE 43

29 Human Services (USDHHS), 1997). Consumption of large amounts of methylmercury-containing fish or seed grai ns treated with methylmercury or other organic mercury compounds may cause perman ent brain and kidney da mage, as well as damage to the brain of a developing fetus. Methylmercury exposure is most dangerous to young children because a greater amount of th e chemical passes into the brain and interferes with brain development. Adding to the problem is that since methylmercury is concentrated in fish muscle tissue (unlike organic contaminants like PCBs and dioxins, which concentrate in the skin and fat), merc ury cannot be filleted or cooked out of consumed fish (USGS, 1995).

PAGE 44

30 CHAPTER 3 PREVIOUS MODEL STRUCTURE Mercury fate and transpor t within the Carson River are simulated using three computer models: RIVMOD (RIVer MODule ), WASP5 (Water Quality Analysis Simulation Program), and MERC4 (MERCury). These models were originally chosen and linked by Heim and Warwick (1997). Curre ntly, RIVMOD is a subroutine within the WASP5 code and is executed during each model time step. Sediment transport is simulated within the WASP5/MERC4 system of models and not within the RIVMOD code. The resulting model framework is referr ed to as the Integrated Fluvial Mercury Transport and Fate Model, or IFMTFM. Hydrodynamic Modeling Using RIVMOD RIVMOD (Hosseinipour and Martin, 1990) is a USEPA one-dimensional (1D) hydrodynamic and sediment transport routine th at simultaneously solves the standard fluid equations of continuity and moment um throughout the system being simulated, utilizing the Gauss-Seidel iterative scheme. Finite difference equations are solved by the Newton-Raphson method to determine fluid velocity and depth given unsteady flow conditions. RIVMOD’s input re quirements include bed elevation, cross-sectional geometry, roughness coefficients, and initial conditions for water depth and discharge. Discharge is specified at the upstrea m and downstream boundary. The model domain includes the Carson River from the USGS Carson City Gage (CCG ) through Lahontan Reservoir. The 115-km, or 71.4 mile, reach is subdivided into 273 segments. The location of each model segment is specified by model stations located at both the upstream and

PAGE 45

31 downstream ends of the segment, resulting in a total of 274 stations. Station 1 is located at CCG and Station 274 at the base of Lahont an Dam. Station spaci ng is generally 0.50 km throughout the modeled reach, although sta tion spacing is reduced to 0.25 km in the transition region between the river and reservoir for stability reasons. Discharge is specified at the upstr eam boundary based on CCG data. At the downstream boundary, discharge from Lahontan Rese rvoir is specified with data from the USGS gaging station below the dam (LDM #10312150). Inflow from the Truckee Canal into Lahontan Reservoir is sp ecified with data from th e USGS gaging station on the Truckee Canal near Hazen, Nevada (TRC #10351400). Data availabl e from the USGS gage at Fort Churchill (FCH #1031200) is used to define a time-varying lateral outflow at this location (IFMTFM station 128); the flow difference between what is predicted and what is measured at FCH is removed, ther eby accounting for irri gation diversions and loss/gain due to groundwate r interaction between th e upstream and downstream boundaries. Here it is important to note th at no diversion ditche s are in operation downstream of the FCH gaging station. Two major modifications have been ma de to RIVMOD. The original RIVMOD code was capable of modeling sediment trans port, but geometry was limited to a simple rectangular channel or a fi rstor second-order polynomial. The first modification implemented a more complex cross-sectional geometry. Figure 3-1 presents the channel representation incorporated in to the modified version of RIVMOD (Warwick and Heim, 1995). Note the cross-section is divided into two basic components: the main channel (parameter BW2) and floodplain (d epth greater than D2). The i nner channel and low-medium flow transition slope portions of the cross-section represent the channel bed.

PAGE 46

32 The portion of D2 above the low-medium flow transition slope represents what will be referred to as the channel bank; the modeled portion of the Carson River is characterized by steep, nearly-vertical channel banks. The RIVMOD input file includes values for each parameter at each model station, the combined result of averaged field measurements and approximate values as taken from USGS t opographic maps. The modified version of the code is able to consider changing flow and cr oss-sectional area between adjacent stations. The second modification applied a divided ch annel approach to the momentum equation, thus allowing for a separate floodplain r oughness coefficient and the computation of overbank depths and velocities (Carroll, 1999) A smaller modification was also made to allow for the inclusion of evaporative losses, which are potentially significant in the semi-arid region of the Carson River system. ` Figure 3-1. Modified channel geometry

PAGE 47

33 The WASP5 and MERC4 Programs WASP5 is the USEPA Water Quality Anal ysis Simulation Program-5 that was developed to simulate the transport and tran sformation of various water body constituents (Ambrose et al., 1991). Mass balance equati ons account for all material entering and leaving model segments through direct and diffuse loading, advective and dispersive transport, and any physical or chemical transformation. Input pa rameters required by WASP5 include simulation and output contro l, model segmentation (same as RIVMOD), advective and dispersive tran sport variables, boundary con centrations, point and diffuse source waste loads, and initial conditions. Several modifications have been made to the WASP5 numeric code. The RIVMOD model is currently a subroutine within th e WASP5 program. This linkage allows for sediment transport information, as relate d to hydrodynamic characteristics, to be incorporated into the WA SP5 mass flux calculations (H eim and Warwick, 1997). The current WASP5 model simulates the transport of three distinct solid types: washload (diameter < 0.063 mm), coarse suspende d sediment, (diameter > 0.063 mm), and bedload. Washload and CSS (coarse suspe nded sediment) are carried in suspension, while bedload is defined as coarse material that travels by rolling, skipping, and/or sliding along the riverbed. Washload a nd CSS boundary conditions are modeled by making use of measured rating curves at CCG ; by contrast, bedload is modeled using van Rijn’s transport theory (van Rijn, 1984a ; van Rijn, 1984b; Carroll, 1999). USGS data indicate that CSS is held constant between CCG and FCH. While exchange is modeled between water column and bed segments, there is no net increase in CSS. However, data also show that on average 70% of the bank material is considered coarse (Heim and Warwick, 1997). Thus, for the purpose of mode ling the system, it is assumed that eroded

PAGE 48

34 coarse bank material settles onto the bed se gment and is transported as bedload. The exchange between bedload and CSS is m odeled by satisfying the water column CSS concentrations as defined by the measured ra ting curves. Equations governing the erosion of washload were originally developed by Ca rroll (1999). Modificati ons to these erosion equations and additional relationships for modeling overbank deposition during extreme flood conditions were later introduced by Carroll et al. (in press). It is assumed that the river is in vertical equilibrium with erosi on and/or deposition occu rring only laterally across the floodplain and not along the cha nnel bed. Channel bank and floodplain are subject to dynamic erosion processes within the IFMTFM, while channel bed is assumed to remain constant. It is further noted that channel geometry is not updated within the RIVMOD code following mode led sediment erosion. MERC4 is a subroutine contained within WASP5 (Martin, 1992). It was developed to compute mercury speciation and kineti c transformation. MERC4 is capable of simulating up to four mercury species: elemental Hg(0), Hg(2+), inert Hg, and Methylmercury (MeHg). The MERC4 model is also capable of simulating up to three solids types: washload, CSS, and bedload in this case. For this study, only inorganic mercury and methylmercury are modeled. Note that only the soluble forms of these mercury species are available for chemical transformation. The simulation of solids is critical because of the strong tendency of mercury to sorb onto solids. The mercury speciation and kinetics algorithms are ba sed on the Mercury Cycling Model (MCM) developed by Tetra Tech, Inc. (Hudson et al., 1994) for the Electric Power Research Institute. The MERC4 model resu lted in the coupling of chemical speciation and kinetics,

PAGE 49

35 based on the MCM model, with the trans port capabilities of the WASP4 modeling system. Within the MERC4/WASP5 program, mass balance equations are written for each form of mercury and solids being modeled. Th e initial total concen trations for each of these state variables are specified to the m odel. New total concentrations are computed based on transport, transformations, and r eaction kinetics, which vary depending on the species present (Martin, 1992). In addition to the state variables, other chemicals which affect mercury speciation are input to the model but not predicted, including the following: pH and concentrations of free sulf ide, sulfate, chlori de, dissolved organic carbon, and dissolved oxygen. The chemical species considered by MERC4 include dissolved, sorbed, and precipitated forms. Measured rating curves at CCG for inor ganic mercury and methylmercury in the water column were written into WASP5 as an upstream boundary condition (Carroll, 1999). Reaction constants for sorption (Bonzongo et al., 1996a); and methylation and demethylation reactions (Carroll, 1999; Carro ll et al., 2000), which are input into the model, are based on field data collected along the Carson River system. Initial conditions for mercury bed concentrations are descri bed by fitting a three-point moving average through data collected by Miller and Lechler (1 998). Initial concentra tions for mercury in the banks were based on calibrated paramete rs and were not dynamic in nature. The previous version of WASP5/MERC4 does not s imulate diffusion other than from the bed sediments. These shortcomings of the model in terms of simulating field conditions will be addressed later in this chapter.

PAGE 50

36 Stages of Model Development The modified and linked RIVMOD/WASP5/ MERC4 model, as described above, makes up the framework of the IFMTFM. M odel development has occurred in three distinct, but ongoing phases (H eim and Warwick, 1997). The first phase of the project applied RIVMOD to the Carson River and Lahontan Reservoir system (CRLR) to simulate a variety of river flows and rese rvoir stages (Warwick and Heim, 1995). The second phase incorporated sediment transport and characterization of the association of mercury with sediment transport. The third pha se seeks to incorporat e the results of the first two phases with MERC4 to determine ar eas of potential mercury methylation, define important transport mechanisms, isolate sources or times of increased mercury contamination, and allow applic ation of the model to differe nt hydrologic scenarios. A working model may then be used to test the outcome of mitigation strategies, such as the removal of mercury sources. This effort may ultimately lead to an effective way to reduce mercury concentrations in the water column, and ultimately the food chain. While much of this research was prev iously completed (Heim and Warwick, 1997; Carroll, 1999; Carroll et al., 2000; Carroll et al., in press) this thesis improves the capabilities of the IFMTFM through the incorporation of bank elements. The purpose behind the development of the bank elements to be integrated into the current IFMTFM code is to develop a means to simulate the spatial and temporal variability of methylmercury within the channel banks of the Carson River system by better simulating the actual physical and biological pro cesses which are occurring in the field. Methylmercury concentrations in the bank are not spatially and tem porally constant but are instead a function of mois ture history. A modified Carson River model is developed to include the relation between flow histor y and moisture levels observed within the

PAGE 51

37 banks. Zones of varying moisture impact me thylation and demethyl ation kinetics, and therefore the temporal concentrations of me thylmercury available to the water column.

PAGE 52

38 CHAPTER 4 CHANNEL BANK ELEMENTS: LINKING SURFACE WATER AND GROUNDWATER Modeling the Carson River The USEPA designated the Carson River, Ne vada as part of a Superfund site in 1991. Between 1860 and 1890, elemental mercury was used at over 70 mill sites to recover gold and silver via mercury amalga mation, introducing 7000 tons of mercury to the Carson watershed. Contaminated sediment s line channel banks downstream of these mill sites, with mercury concentrations two to three orders of magnitude above background levels. Mercury methylation within sediments, accompanied by diffusion and erosion, introduces methylmercury to the cha nnel. Methylmercury concentrations in the bank are not spatially and tem porally constant but are inst ead a function of moisture history. A modified Carson River model is deve loped to relate flow history to moisture levels observed within the banks. Zones of varying moisture impact methylation and demethylation kinetics, and therefore the te mporal concentrations of methylmercury available to the water column. Methylmercury concentrations within the Carson River channel banks vary with both spatial location and time. The temporal va riability in methylmercury concentrations was not previously considered by the IFMTFM The previous modeling approach (prior to this thesis) considered me thylmercury bank concentrations as temporally constant and inversely related to the square root of channel bottom slope (S0) (Carroll, 1999):

PAGE 53

39 (Eq. 4-1) A calibrated threshold flow of 1000 cfs was developed, below which methylmercury is available to be eroded from the channel banks. This threshold flow was later modified to a threshold depth of 1.00 meters. Diffusion of methylmercury was only allowed to occur from the channel bottom of the main channel. No efforts are included within the previous model to simulate groundwater hydrodynamics in response to flow history within the river channel. It is this tem poral variability in bank satu ration zones that controls the kinetics underlying the microbial production of methylmercury within the channel banks, thereby influencing MeHg available for in troduction to the water column. Research concerning the relation between soil wetti ng and drying cycles; and methylation / demethylation kinetics is ongoing at other academic institutions. The end product of this thesis research is the introduction of ch annel banks into the IFMTFM code as areas for active mercury transformation. Channel bank elements are created within the IFMTFM, conceptually sim ilar in functionality to bed elements within the IFMTFM. Bank element size is dynamic and equal to the saturated portion of the bank. Methylation and demethylation will o ccur within channel banks, resulting in dynamic methylmercury concentrations. As in the previous IFMTFM version, channel banks are eroded, but the methylmercury concen trations within the eroded material are no longer temporally constant. The code was al ready able to predict flow magnitudes and depths along the Carson River from th e Carson City Gage through the Lahontan Reservoir. The creation of bank elemen ts, along with the accompanying coding modifications, allows the model to predict the zone of saturation within the channel banks. The functionality of these bank elemen ts is determined based on analysis of the 5 0 0 2] [ S HgMeHg

PAGE 54

40 following: river stage predictions as com puted from a 3-dimens ional (3D) hydrodynamic model, water table data from three groundw ater wells, and the time lag between river stage oscillations and observ ed water table response. Available Data To obtain more precise channel geometry data, a GPS (global positioning system) unit, the Leica GPS System 500 (Leica Geosys tems, Atlanta, Georgia), was purchased and used to collect detailed data for a 0.5 km reach of the lower Carson River, with this reach encompassing the location three instal led groundwater wells (Scout Camp field site). Signals from at least four satellites are needed by a GPS rece iver on the ground to resolve measurement ambiguities and obtain the location (latitude, longitude, and elevation) of a given observed point. Without getting into the comple xities of the system, observations from four satellit es are sufficient to resolve ambiguities and obtain spatial location on the surface of the earth. The repe tition of this procedure allows for the collection of detailed channel morphology data for this small reach of the Carson River. Vertical resolution of data points, based on syst em specifications, is re ported to be on the order of a few centimeters. Flow data (magnitude and depth) from the USGS gaging station at Fort Churchill, water table data (March 2001 through presen t) from the three wells installed along a channel transect, and GPS survey data are us ed to construct a de tailed surface-water model and to integrate channel and groundw ater hydrodynamics. Spec ifically, the goal was to develop simple relationships between flow history (specifically rising and falling water depths) in the channel to moisture c ontent values (specifically the height of saturation) within the channel banks. The si mple relationships discovered from modeling efforts and data analysis were incorporated into the IFMT FM modeling of channel bank

PAGE 55

41 elements. With the IFMTFM model then able to predict bank moisture content both spatially (due to variable channel geometry ) and temporally, microbial kinetics data was included to allow the model to predict spa tial gradients and tem poral variability of methylmercury concentrations within the ba nks of the Carson River. The inclusion of bank elements allows the IFMTFM to more accurately simulate mercury concentrations present in eroded bank material and methyl mercury gradients for diffusion processes. Calibration and verification of the model is used to compare the updated IFMTFM model with data collected within the Carson River system. Although the bank element is conceptually simple, a historical perspectiv e with regard to mode ling of surface water and groundwater interaction is important to ga in additional insight prior to a discussion on data analysis. Introduction to the Hyporheic Zone Historically, much ecological and hydrological research on groundwater and rivers has treated the two systems as distinct enti ties. River and groundwater systems are, in fact, characterized by marked differences (B oulton et al., 1998). Ri vers typically have currents generating turbulence; short water residence time; variable discharge and physicochemical conditions; dynamic channel morphology; unidirectional transport of nutrients, sediments, and biota; and are open to sunlight. By contrast, alluvial groundwater environments tend to offer increas ed stability, longer water residence times, laminar flow regimes, little change in sediment bed structure, and permanent darkness. Despite these differences, the interface betw een surface streams and alluvial groundwater represents a region of active in terchange of water, nutrients, biota, and other materials. This intervening zone is termed the hypor heic zone. Exchange occurs because of variations in discharge, bed topography, a nd porosity (Boulton et al., 1998). Upwelling

PAGE 56

42 groundwater supplies nutrients to stream or ganisms, while downwelling stream water provides dissolved oxygen and organic matte r to microbes and i nvertebrates in the hyporheic zone (where photoaut otrophic production is not po ssible). Dynamic gradients exist at all scales (due la rgely to sediment heterogene ity), and the bound aries of the hyporheic zone vary in both time and space. Flow occurs in three dimensions: longitudinally down the channel, vertic ally between the ch annel and underlying sediments, and laterally to or from th e riparian zone (Jones and Holmes, 1996). Subsurface regions are classified as hyporhe ic or groundwater zones based on hydrologic exchange with surface waters; one definition proposes the division between the hyporheic zone and groundwater to be where less than 10 % of the subsurface wa ter originated from the surface channel (Fuller and Harvey, 2000; Jones and Holmes, 1996). The dynamics of the hyporheic zone are particularly important for mercury speciation. As the extent of the hyporheic zone fluctuates with stream stage and the associated changes in hydraulic gradient s, so too do the anoxic zones where the methylating population is able to thrive. Fl ow rates through sediments are much slower than on the surface, thus constraining th e import of dissolved oxygen. If oxygen demand from aerobic metabolism and other chemical re actions exceeds the influx, anoxic regions will develop and have two important effects on hyporheic zone processes (Jones and Holmes, 1998). First, organic matter decomposition occurs via anaerobic pathways, such as denitrification, su lfate-reduction, and methanogene sis. Secondly, the oxic-anoxic interface is the site of redox/pr ecipitation reactions, such as the oxidation of methane or ferrous iron. Additionally, me thylating SRB thrive in these anoxic environments. The active exchange of water in the hyporheic zone with water in the stream enables

PAGE 57

43 methylmercury introduction to the biota of the channel and offers an additional introduction pathway to bank erosion. The cont ribution of hyporheic zone processes to the stream ecosystem function is determined by the rate of chemical transformations in the sediments and the proportion of stream discharge flowing through sediments (Jones and Holmes, 1996). It should be further noted that flow of surface water in the hyporheic zone increases the effective reactive site density per volume of surface water by providing contact of dissolved metals with potential reaction or sorption sites, thus potentially enhancing metal attenuation by tran sport into zones wher e removal processes are more favorable (Fuller and Harvey, 2000). Modeling Surface-Subsurface Interact ions: A Historical Perspective A number of different appro aches were taken over the years in the effort to quantify stream-aquifer hydraulics. Individual efforts to obtain analytical expressions to relate stream and aquifer hydraulic heads are each based on the same starting point, the continuity equation. The differences arise in the assumptions made and the consideration of different boundary conditions. In the inte rest of completeness, a number of these approaches are summarized within this section. Groundwater levels in alluvi al valley aquifers vary with stage changes in the associated surface streams. A rise or fall of stream stage over a short time period imparts a flow reversal within the aqui fer, a result of the change in gradient between the stream and the aquifer (Serrano and Workman, 1998). The distance of infl uence within the aquifer is dependent on the aquifer transmi ssivity and porosity, the change in stream stage, and the length of time that conditions for a flow reversal exist (Serrano and Workman, 1998). The hydraulic grad ient is also spatially vari able; river stage typically does not increase uniformly over a long reach be cause of variations in channel geometry

PAGE 58

44 and bottom slope (Squillace, 1996). When water quality within the stream differs from that within the aquifer, the spread of contaminants (e.g. methylmercury) is a problem directly connected to the hydraulics of the stream-aquifer system. A rise in river stage, as associated wi th downstream movement of a flood wave, is accompanied by lateral stream losses to bank sediments. As the flood wave passes, the volume of water held in bank storage is gra dually released to sust ain baseflows between flood events. Channel banks and floodplains thus mitigate against extreme hydrologic phenomena by storing substantial volumes of water which would otherwise increase flood volumes, a process known as bank st orage (Whiting and Pomeranets, 1997; Hantush et al., 2001). While this interact ion between surface streams and groundwater has often been studied in the interest of flood control, th e underlying equations remain the same for gradually varied stage changes within the stream. The discussion to follow includes mention of a number of solutions developed for unconfined aquifers. Yeh (1970) and Marino (1975) provide nume rical solutions by accounting for streams as time-varying boundary conditions. Further numerical solutions have been developed modeling the stream a nd aquifer as interacting dynamic systems, including work by Pinder and Sauer ( 1971) and Zitta and Wiggert (1971). One-dimensional open channel flow equations are coupled with the two-dimensional (2D) ground water flow equation. Assumpti ons embedded within these solutions may bring about questions as to th eir practicality. Despite being more restrictive, practical analytical solutions have also been developed. Several author s have presented analytical solutions to the linearized Boussinesq equa tion, subject to a fluctu ating stream stage. Cooper and Rorabaugh (1963) derived an an alytical solution fo r the changes in

PAGE 59

45 groundwater head when the river stage can be approximated as a damped sinusoid, as in the case of a passing flood wave. A finite diffe rence expression of the analytical solution developed by Cooper and Rorabaugh was de veloped by Hornberger (1970). Hall and Moench (1972) and Moench et al. (1974) use the convolution integral to expand applicability to flood waves of arbitrary shap e, as described by a step function. A more general solution developed by Workman et al (1997) includes the e ffects of transient redistribution of water levels in the aquifer from the pr evious day. Govindaraju and Koelliker (1994) assessed the accuracy of th ese analytical solutions compared with numerical solutions of the nonlinear equa tion, finding as much as 10% error. The analytical solutions do not consider the simultaneous interactions that are inherent to a dynamically coupled stream-aquife r system; rather, their use is based on a prior knowledge of the stream-stage fluctuat ions, and they cannot be used for routing streamflows without iterative procedures (H antush et al., 2001). One key source of difficulty in the efficient nume rical coupling of streamflow and aquifer equations is the intrinsic difference in the time scales of wate r movement in the two systems (Perkins and Koussis, 1996). Analytical solutions to simu ltaneously couple the governing equations of stream flow and groundwater flow are deve loped by Morel-Seytoux (1975), Hunt (1990), and Harada et al. (2000). Anal ytical solutions to the non-li near Boussinesq equation are later presented by and Serrano and Workman (1998). Each of the above solutions was developed starting from the following standard assumptions, with additional assumptions included as needed (Sharp, 1977): The alluvial aquifer is homogeneous, isotropic, and infinite in extent. The alluvium’s bottom boundary is horizontal and impermeable. The Dupuit-Forchheimer conditions are valid. These conditions are: o In any vertical section the gr oundwater flow is horizontal;

PAGE 60

46 o The velocity is uniform over the depth of flow; o The slope ( ) of the free surface (unconfined flow) is small enough that tan ( sin ( ). All media are saturated. Water from storage is discharge in stantaneously upon reduction in head. Fully-penetrating streams (i.e., the streams pe netrate to the bottom of the alluvium, or alternatively, the heads in the alluvium ben eath the river bed are all times equal to the river stage. A solely unconfined or solely confined groundwater system. Some authors have sought to develop more general so lutions by relaxing some of these assumptions. Solutions derived through us e of linearization tech niques applied to the nonlinear Boussinesq equation are naturall y expected to contain some error, caused by the nonlinear nature of the original equa tion (Gill, 1985). This being the case, one may choose to use even simpler solutions, using suitably calibrated values of the aquifer constants to obtain reasonable results. G ill (1985) deduced limiting solutions for finite aquifers of long-thin and shortthick dimensions, rather than assuming aquifers of infinite thickness. Hantush (1967) derived 2D, tim e-dependent equations for a stream-leaky aquifer system subject to an instantaneous ri se in stream water level, with a constant water level thereafter. Dever and Cleary (1979) advanced Hantushs approach with general solutions for any stream stage hydrograph that can be represented by a continuous or piecewise continuous function of time. An analytical model of stream-aquifer interaction that considers the effects from a small degree of aquifer penetration and low-permeability sediments on the head response to an arbitrary stream stage hydrograph was proposed by Zlotnik and Huang (1999). The progression of research into the area of stream-aquifer interaction led to the development of numerical models, which became required tools as assumptions were relaxed and solutions became increasingly complex and realistic. Neuman and

PAGE 61

47 Witherspoon (1971) developed an iterative, fini te element method applicable to a wide variety of problems involving complex boundaries and arbitrary degrees of heterogeneity and anisotropy. This approach, which consid ers 2D flow within channel banks (and thereby does not encompass the limitations of Dupuit assumptions), was used in the development of a bank storage model (WaTab2D) by Whiting and Pomeranets (1997). Two-dimensional flow within channel banks for unconfined aquifers, of either semiinfinite or finite width and without a semipe rvious streambank, was further considered in work by Moench and Barlow (2000). It was necessary to assume that water table fluctuations were small relative to the satura ted thickness and that ve rtical flow into or out of the zone above the water table occurr ed instantaneously, therefore neglecting the effects of hysteresis in the moisture distri bution above the water ta ble. In addition to considering stream interaction with anis otropic, homogeneous unconfined aquifers, Moench and Barlow (2000) also considered stream interaction with various homogeneous confined and leaky aquifer types. The computer programs STLK1 and STWT1 were developed for simulations of confined and l eaky aquifers and confined and water-table aquifers, respectively (Barlow et. al., 2000). It was noted that a water-table aquifer (unconfined) may respond very much like a conf ined aquifer if the unsaturated zone is thin and the specific yield is small, the resu lt of the non-instantane ous release of water from storage following a reduction in hea d. Freeze (1971) was one of the earliest researchers to consider the integration of the saturated and unsaturated zones into a transient model. One common failure with the use of non-integrated fully saturated flow models is that slower wate r table responses are predicte d than are actually observed (Wise et al., 1994).

PAGE 62

48 In the choice of a model for a given sy stem, it is important to consider the assumptions upon which the model is based. Assumptions made for some of these solutions are likely to be questionable in so me instances and can yield erroneous results. These questionable assumptions include D upuit-Forcheimer conditions, homogeneity, full penetration of streams, and systems that are solely confined or unconfined (Sharp, 1977). Sharp used data collected from major a lluvial valleys in the central United States to assess the validity of these assumptions; his findings are listed below: The assumption of homogeneity is clearly invalid as demonstrated by the exponential increase in hydraulic conductivity with depth. Isotropy assumptions might be justified in the deep alluvium; anisotropic conditions may occasionally be present at shallower depths. Valley bottoms may be adequately represen ted by a horizontal surface; the extremely high hydraulic conductivities of the deep alluvium also justify the assumption of impermeable boundaries, at leas t for first approximations. It is clearly shown that Dupuit-Forcheimer assumptions are not valid in the major alluvial systems considered. Flow is predomin antly vertical in the vicinity of the river bed; groundwater flow is also not uniform with depth. While the approach reduces the problem’s dimension by one, it does not allo w accounting for the vertical component of the seepage velocity or accurate predic tion of a significantly sloped free surface (Whiting and Pomeranets, 1997). The assumption of saturated flow is legi timate and typical co nductivities allow for very rapid (instantaneous) release of water from storage upon head reduction. The rivers of the major alluvial aquifers studied are not fully pe netrating. Presumably, most groundwater flow between the river a nd bank storage is through the river bed and not the banks because: o The river bed has grea ter surface area, and o The river bed is closer to the zone of great er hydraulic conductivity. Additionally, the banks and bed, in general, will have differing hydraulic conductivity because of smaller sediment si zes and higher clay percents in the bank than in the bed, further decreasing th e contribution of the banks to the river-aquifer interaction. This condition leads, in pa rt, to the invalidity of the Dupuit-Forcheimer

PAGE 63

49 assumptions and makes most 1D models suspect. The simulation of bank storage using semi-permeable stream banks is not directly applicable. Alluvial aquifers may be confined and unconfined and these conditions may vary both spatially and temporally. Based on the above discussion, it becomes a pparent that many bank storage models are based on assumptions that do not corre spond to observed field conditions. As noted by Sharp, this does not necessari ly invalidate the models. Rather, the validity of a model and its assumptions rests on two divergent sets of criteria: “The positivistic view states that any assumption which leads to a useable predictive model is a valid assumption…the assumptions are valid regardless of any co rrespondence to observ ed or theoretical processes. The normative view states that assumptions should co rrespond to the actual processes or the true relationships between va riables. If not, the assumptions are invalid because they yield or are faulty interpreta tions” (Sharp, 1977). The listed assumptions fall into both classifications, with difficulty arising when positivis tic assumptions are used to infer the actual hydrogeologic processes or when data from a positivistic model is extrapolated to a situation where conditions differ by any significant degree (Sharp, 1977). “The problems involved in modeling without simplifying assumptions are self-evident. Hydrogeologic va riability insures that models cannot precisely correspond to real-world processes. Nevertheless, the hydrologist or geologist who is attempting to describe or model and bank-storage situati on must be aware of the limitations of his assumptions. The degree to which invalid (in the normative sense) assumptions reduce the accuracy or interpretation or prediction will depend on the degree of divergence from the real world and cannot be pr edicted a priori” (Sharp, 1977).

PAGE 64

50 Modeling Simplifications Despite the probable presence of a regional groundwater flow perpendicular to the Scout Camp groundwater well transect, limiti ng data requires the assumption that water fluxes along the front and back boundaries of this cross-section are negligible in comparison to fluxes along the cross-section (perpendicular to the stream). River stage and water table levels are then able to be analyzed as a 2D model. It is essentially assumed that changes in hydraulic head with in the aquifer are due primarily to the pressure head gradients associated with the changing river stage, rather than changes in aquifer head levels further upstream. This phenomenon is shown schematically in Figure 4-1. Using IFMTFM Modifications to Trac k Bank Saturation: Bank Element Functionality Knowledge in hand, the functi onality of the IFMTFM m odel is next considered; specifically, how the model pr eviously operated is compar ed with the updated model Figure 4-1. Water table res ponse to changing river stage River stage

PAGE 65

51 with incorporated bank elements. Sediment transport and mercury transport and kinetics within WASP5/MERC4 are reviewed. The pr evious IFMTFM includes two types of elements: water column elements and bed elements. The modeling of mercury cycling within channel banks required the creation of a new type of element: the bank element. The bank element is used to track moisture content levels within the channel banks. As moisture contents change within bank elemen ts, properties of these elements relating to mercury modeling are updated. The ability to model moisture content variability within the channel banks, utilizing the newly-deve loped bank elements, allows for a more detailed analysis of microbial behavior concerning methylatio n and demethylation. Mercury modeling within the previous IFMTFM was accomplished as follows (Carroll, 1999). The Carson City Gage (CCG) measured rating curves for washload and coarse suspended sediment, along with a cal culated bedload rating curve, are written directly in the WASP5 code as upstream boundary conditions, or ra ther, input to the model domain. Measured rating curves for i norganic mercury and methylmercury in the water column are also directly written into the WASP5 code as upstream boundary conditions. Mercury concentrations within the channel banks we re chosen to calibrate the water column concentrations of inorganic mercury and methylmercury. Miller et al. (1999) found a statistically sign ificant correlation between mercury concentrations and the percentage of fine sediment (washloa d, diameter < 0.063 mm), with the implication being that the contaminated material is primar ily associated with the fine sediment. This fine sediment is deposited along the lower re aches of the Carson Ri ver where the stream gradient is shallow and velocities are slow The assumption was therefore made that

PAGE 66

52 concentrations of inorganic mercury and me thylmercury within the bank sediments are inversely related to the slope of the channel: 5 0 0 1] [ S Hgbank in (Eq. 4-2) 5 0 0 2S Hgbank MeHg (Eq. 4-1) S0 is the channel bottom slope and 1 and 2 are calibration parameters adjusted to match water column mercury concentration data. Bank mercury concentrations (inorganic and methylmercury) are not dynamically changed in the previous version of the model. This IFMTFM version includes a threshold flow depth (based on segment geometry and calibrated QT = 1000 cfs) above which methylmercur y does not occur in the channel banks because methylating microbial communities cannot survive. Within the confines of the previous model, methylati on and demethylation (resulting in dynamic concentrations of MeHg and inorganic Hg(II)) are allowed to occur in the channel bed sediments (confined by the main channel) and th ese kinetic reactions are governed by the equations 4-3 and 4-4: ] [ ] [20 r temp pC K f t C (Eq. 4-3) 10 20 10 T tempQ f (Eq. 4-4) For mercury methylation, the reactant (Cr) is soluble Hg2+ and the product (Cp) is soluble methylmercury. In the case of demethylat ion, soluble methylmercury serves as the reactant and soluble Hg2+ serves as the product. K20 is the reaction rate (day-1), ftemp is a temperature multiplier to correct for temper atures deviating from 20C (dimensionless), t is time in days, and reactant and product concentrations are in units of milligrams per liter

PAGE 67

53 (mg/L). The spatial variability of the constants K20 and Q10 are summarized in Table 4-1 (Carroll, 1999; Carroll et al., 2000). Table 4-1. Spatial variability of reaction constants within IFMTFM Range MethylationDemethylation (distance from CCG in km) K20 (day-1)Q10K20 (day-1)Q100.00 79.25 0.00412.030.44832.32 79.25 95.400.00112.031.32912.32 95.40 115.000.00282.031.17532.32 Initial mercury concentrations in bed sediments were determined by fitting a three-point moving average thr ough data collected by Miller and Lechler (1998) (Carroll, 1999). Inorganic mercury and methylmercury were introduced to the water column through the following mechanisms (Figure 2-2) : erosion of channel bank (vertical portion of channel at BW2 in Figure 3-1) sediment s; and diffusion from bottom sediments within the confines of the main channel (BW2 in Figure 3-1) (inner channel is assumed in dynamic vertical equilibrium and therefore not subject to erosion). Only deposition, not erosion, is considered to o ccur during flows which contac t the floodplain (BW2 to BW3 and BW3 to BW4 in Figure 3-1). Mercury conc entrations within the eroded sediments remain constant over time, although concentrat ions are spatially va riable based on the 1 and 2 calibration parameters. Water column concentrations are based on measured partitioning constants between dissolved and adsorbed phases. MeHg and inorganic Hg(II) are the only species modeled, based on the high volatility of elemental mercury. Methylation and demethylation only occur w ithin the water-contacted bed sediments of the main channel, based on the predicted fr ee ionic MeHg and free ionic inorganic Hg(II) concentrations. Free ionic (~soluble) concen trations are considered to represent the bioavailable fraction of total mercury, with so luble fraction defined as the fraction able to pass through a 0.45 m sieve. Partitioning coeffici ents are shown in Figure 4-2.

PAGE 68

54 Figure 4-2. State variables measured with MERC4. (Reprinted with permission from Carroll, R.W.H., 1999. Simulation of Me rcury Transport and Fate in the Carson River, Nevada. Masters Thesis University of Nevada-Reno, Reno, Nevada. Figure 1.4, pp. 16) Based on these considerations, the previ ous IFMTFM version was calibrated to provide a best-fit to field observations of water column mercury and sediment concentrations. This approach, however, is cl early not entirely repr esentative of actual field processes. The calibrated methylmerc ury function predicts methylmercury bank concentrations over two orders of magnitude higher than the methylmercury bank data collected under conditio ns of near-zero flow (Carroll, 1999). The samples, from October 1994, were collected high up on the bank at the estimated point of ri ver contact during spring floods. Examination of water column data shows that, unlike other systems, increased discharge does not always cause a dilution in MeHg. Instead, an increase from low to medium flow causes a significant incr ease in MeHg water column concentrations. These results strongly suggest that given medium flows, bank erosion, and not diffusion, acts as the principal pathway for MeHg into the water column. This trend of MeHg increase with increasing flow begins to reverse when flows go from medium to high conditions. The resulting diluti on at higher flows suggests that the availability of MeHg

PAGE 69

55 from erosion of the upper bank is limited, a nd this observation was used to justify the inclusion of a threshold height got bank Me Hg in the previous model version. This limitation is not observed for inor ganic mercury (Carroll, 1999). Modeling MeHg concentrations as constant ov er time is not repr esentative of field conditions, which depend on mi crobe populations and therefore moisture levels. Mercury methylation is stimulated if soils are fully saturated and anoxic conditions are achieved (Gilmour and Henry, 1991; Porvari and Verta, 1994). Low flow conditions, which characterize the Carson River during most of the year, do not result in inundation of channel banks or the attainment of anoxic c onditions. It is probable that high flow rates are too infrequent and short in duration to allow anaerobic microbes to develop in the upper portions of the bank sediments. Given th ese considerations, MeHg concentrations within the channel banks are not expected to be vertically uniform; rather, MeHg concentrations are expected to display a de pendence on soil moisture conditions (Carroll, 1999). Flow history and the rates at which river stages are risi ng or falling will, in turn, impart time-dependent and vert ically variable MeHg concen trations within the channel banks along the Carson River. Th is hypothesis is borne-out in soil samples collected at Deer Run Road. Four soil samples were collect ed in triplicate from the bank profile at this location, with sampling locations show n in Figure 4-3(A). Evaluation of sample water content and percent methylmercury resulted in the relationship shown in Figure 4-3(B), which clearly shows a direct relati onship between the percent of MeHg in the bank profile and the soil mois ture content (Carroll, 1999). As mentioned previously, the previous version of IFMTFM utilizes temporally constant MeHg bank concentrations and a threshold fl ow depth above which MeHg does not occur

PAGE 70

56 in the channel banks, both of which were de termined through calib ration procedures. The incorporation of the bank element into the IFMTFM attempts to remedy this and other discrepancies between the model and physical observations; and reduce the number of calibration parameters. Modeling efforts us ing the Finite Element Surface-Water Modeling System: Two-Dimensional Flow in a Horizontal Plane (FESWMS-2DH) model and analysis of water table data from the installed groundwater m onitoring wells allowed for the determination of a temporally-varia ble zone of saturati on (as critical for methylation) within the channel banks of the updated IFMTFM version. The zone of saturation is a function of tim e history of channel stage, as computed within RIVMOD. The saturated extent of channel banks is also a function of spatial location within the model domain, resulting from variable channe l geometry, and hence computed stage, for any given flow magnitude. The creation of the new bank element type within IFMTFM allows for moisture content within the banks will be continuously simulated and updated as required throughout the model domain from CCG to Lahontan Reservoir. Within the domains of RIVMOD/WASP5/M ERC4, each bank element includes the channel bank between adjacent stations (C arson River model includes 274 stations). Recall that different channel geometry is assi gned for each station within the input files. This geometrical configuration inside each m odel segment (between adjacent stations) is determined through linear inter polation of the properties of the two bordering stations. Symmetry is specified for opposite channel banks. A single bank element is therefore sufficient for each IFMTFM model segment. Flow history must be tracke d. The combined vertical exte nt of the capillary fringe and the partially saturated zone will depend on whether river stage is rising or falling. A

PAGE 71

57 Figure 4-3. Dependency of methylmerc ury concentration on soil moisture. (A) Schematic of bank profile at DRR with sa mpling locations marked. (B) Percent MeHg versus percent moisture content found in the bank profile at DRR. (Reprinted with permission from Carroll, R.W.H., 1999. Simulati on of Mercury Transport and Fate in the Carson River, Nevada. Masters Thesis. Un iversity of Nevada-Reno, Reno, Nevada. Figure 1.13, pp. 38) smaller distance would be expected if the river stage is rising (bank not recently inundated) and a larger distance if river stage is falling (bank recently inundated). This is a by-product of the variable relati onships between pressure head ( ) and moisture content ( ), based on whether a given soil is wett ing or drying. This relationship is commonly referred to as hysteresis and can be represented by scanning curves (Tindall and Kunkel, 1999). A B

PAGE 72

58 Two soil samples were collected along the channel bank at the location of the groundwater wells during Octobe r 2002. While possible to obta in a steady-state estimate for the size of the capillary fringe using these samples, incorporation of this value into the IFMTFM would require the assumption of equ ilibrium attainment within bank sediments. Given the nature of Carson River streamflow equilibrium conditions will rarely be reached (except perhaps during flood events). Ideally, data would be collected relating the extent of the capillary fringe to the rates at which channel stage is rising or falling. Realistically, this data collection is impr actical and, given the sp atial heterogeneity of sediments within the Carson River system unilaterally applying these functional relationships to all bank sediments within th e model domain is incorrect. Alternatively, it will be kept in mind that a capillary fringe does exist, that MeHg concentrations within channel bank sediments are likely underestim ated, and, by neglecting th is fringe, that the final calibration to match observed MeHg water column concentrations will be conservatively low. Because mass failure al ong channel banks is not a modeled erosion process within IFMTFM, mercury enters the water column only through erosion or diffusion from contacted (by st ream stage) portions of th e bank. Mercury concentrations within the capillary fringe will only impact the model if these portions of the bank become contacted in the immediate future. Relative to the cont acted portion of the channel bank, the extent of the capillary fri nge is predicted to be small. With these considerations in mind, attempts are only be made to track the saturated portion of the channel bank elements. Soil heterogeneity, the lack of available da ta regarding soil properties relating to soil moisture, the presumed minimal impact of the capillary fringe on water column

PAGE 73

59 mercury concentrations, and data analysis suggesting a nearly instantaneous response time from channel to wells ultimately pushed modeling efforts away from the inclusion of unsaturated zone modeling. The possibility of a variable time lag, from river stage change to water table response, based on whether the bank sediments are wetting or drying, however, was continually considered throughout the data analysis portion of this project. By tracking the size of the saturated zone over time as river flows change, the updated IFMTFM begins to remove itself from the unrealistic assumption of static MeHg concentrations within the ch annel banks and no MeHg above a given depth threshold, as present in the previous model version. By allowing the model to run for a sufficient length of time prior to generating used output initial conditions (bank concentrations) effectively wash out of the system, there by removing the dependence of output on initial conditions. The addition of the bank element allows for user speci fication of different methylation and demethylation rates within the channel banks, as compared to the channel bed. This is expected based on the higher percentage of fi ne particles within channel bank sediments. Data has been pr esented to suggest that methylation is dependent on moisture content and, specifica lly, the presence of a noxic regions within the channel banks. Ongoing resear ch at other institutions is seeking to determine these rates as a function of soil moisture content and the time duration of saturation conditions necessary for maximum MeHg production by micr obes; preliminary results are used in this thesis research. As moisture content within the bank elemen ts change over time, the concentrations of methylmercury with in the channel banks become dynami c, variable with respect to

PAGE 74

60 time, spatial location within the model dom ain, and vertical location within the bank. Variable concentrations are then used in the model to determine methylmercury amounts introduced to the water column via erosion, diffusion, and advecti on (during receding hydrographs) from the banks. Note also that while erosion of channel bank MeHg and inorganic Hg(II) was previously simulated by the IFMTFM, diffusion of these chemicals from channel banks is an addition to the c ode. The inclusion of bank diffusion could be a significant contributor to MeHg concentrations in the water column under medium to low flow conditions. Diffusion was previously onl y modeled from the bed sediments, which include a much larger contact area for diffu sion than do the channel banks. However, mercury concentrations within the channel ba nks are observed to be two to three orders of magnitude greater than the concentra tions within the channel bed sediments. Additional considerations were also needed to determine the impact of moisture content history on methylation and demethylation ra tes. Although the rate s are a function of moisture content, the microbially-mediated nature of these chemical reactions does suggest that response time for each microbial community should be considered and somehow modeled. For example, how long doe s a portion of the bank need to be wet before the observed net methylation rate reflects this moisture change? These considerations are later discussed in ch apter 7. Additionally noted is that water temperature, which impacts the ki netics, is known by the model. Final Considerations and Comparative Analysis The completion of modeling efforts allows for an improved understanding of the relationship between river stage change a nd water table. Based on the high degree of spatial heterogeneity within the system, the simple relationships developed for the Scout Camp reach are unlikely to precisely duplic ate the field conditions observed elsewhere

PAGE 75

61 within the Carson River system. The detail ed FESWMS-2DH and groundwater data were only tools used in developing the relationships between flow history and bank saturation levels. The IFMTFM simulates the hydrodynamics within the system based on user-input data from the USGS gages and through application of RIVMOD. RIVMOD is capable of predicting the temporal and spatial variability of river stage. The water table observations are not known to the IFMTFM model, as the groundwater wells were not installed during periods of historical data collection (used for model calibration and verification). Modified input data file st ructure and a new portion of FO RTRAN code are added to the IFMTFM to incorporate bank elements a nd dynamically compute and update MeHg concentrations in the channel banks. With the bank element concept presented to the reader, it is possible to move forward to the model development and data anal ysis portions of this thesis. Following the discussion of the data modeling and data anal ysis, the final component of this thesis research is a comparative analysis to compare RIVMOD and FESWMS-2DH (SMS) predictions for river stage and flow velocity as functions of flow magnitude. Ideally, the results would be in perfect agreement. Regardle ss of the results, it mu st be realized that the IFMTFM will still only have available to it the stage output from RIVMOD (a 1D hydrodynamic model based on coarse, simplified channel geometry). However, the potential use for higher resolution channel geom etry data (for future model inclusion) can be assessed. Bank elements are used to model the sp atial and temporal variability of bank moisture conditions (over a range of flow conditions and scenarios) to determine bank MeHg concentrations; these bank concentratio ns are then the determining factors for

PAGE 76

62 water column concentrations and, ultimately, the availability of Me Hg for uptake within the aquatic food chain. The development and inclusion of bank dynamics within the IFMTFM will assist in the modeling efforts for the mercury-contaminated Carson River system, with the long-term goal of assisti ng in the determination of an appropriate remediation strategy. Should additio nal data regarding methylati on kinetics as they relate to soil moisture content become available, the basic IFMTFM code structure needed for its inclusion is in place following the completion of this thesis research.

PAGE 77

63 CHAPTER 5 SURFACE WATER MODELING Site Selection The starting point for all field research is site selection. In the case of this study of the Carson River, one of the goals is the quant ification of the interaction between surface water and groundwater. Months in advance of the first field expe dition, it was decided that this would be accomplished through a combination of surface water modeling and analysis of data collected fr om near-channel groundwater well s. Site selection therefore hinged on two critical elements: the ability to gain vehicular access to the site for the survey equipment and the drill rig for well installation, and permi ssion of the landowner to install the wells. Additionally, the site needed to have the characteristic vertical banks and a site layout conducive to and stable e nough for getting a drill rig as close to the channel as possible. Much of the land along the lower Carson Ri ver is privately owned by cattle farmers and agricultural interests, la ndowners with no real reason to justify access to their land and not interested in suppor ting a water quality study. Fortuna tely, a portion of the lands adjacent to the lower Carson River are public lands under the administration of Fort Churchill State Historic Park of the Nevada Division of State Parks. On March 5, 2001, accompanied by Mr. James Prida of the Fort Ch urchill State Park, a suitable site was located at a Boy Scout Camp approximately 10 km downstream of the Fort Churchill USGS Gaging Station (Figur e 1-2). On March 8, 2001, th ree groundwater wells were installed along the south bank of the Carson Ri ver at the Scout Camp field site. The three

PAGE 78

64 wells were installed along a tr ansect, perpendicular to stream flow direction, at distances from the vertical channel bank of approximate ly 5, 25, and 100 feet. The 6-inch diameter, PVC wells are cased to a depth of 20 feet. Pr essure transducers monitor the position of the water table within each well, data which is recorded by data loggers at 1-hour sampling intervals. Collection of GPS Survey Data The channel geometry used in RIVMOD can be described as coarse, based only on outdated USGS topographic maps. The construction of a surface water model required more precise channel geometry data. The Leica GPS System 500 was used to conduct a detailed bathymetric survey of the Carson Ri ver channel for a reach centered on the three installed groundwater wells. The system utilized a base station and a rover unit. Survey data was collected over two periods : March 6-9, 2001 and July 2-6, 2001. The complete set of surveyed elevation da ta points is shown in Figure 5-1. Data points could be divided into several classifications: tran sects (100 through 500), top of vertical banks, water surface, channel thalwe g, or general channel bed points. Surveyed elevations are expected to be accurate to within +/3.0 centimeters (cm), with horizontal accuracy closer to +/1.0 cm The total surveyed reach measured approximately 500 meters along the thalweg. The bathymetric data collected was used as the framework for a surface water model. As the raw surveyed data will later be refined, a few shortcomings of the survey process are herein noted. Recall that GPS is based on travel times and distances from receiver to at least four satellites. The la rge canopy of the cottonw ood trees situated along the channel banks often caused a loss of signal with the requir ed satellites, resulting in incomplete data along the top of the channel banks and effectively limiting the ability to

PAGE 79

65 Figure 5-1. Spatial location of survey point s and cross-sections at Scout Camp site collect precise eleva tion data for the floodplain. Vertic al banks between two and three meters in height additionally limited the ability to collect elevation data adjacent to the base of the vertical banks. Observed st reamflows during March ranged from 165 to 215 cfs, while those in July were between 8 and 9 cfs. Given that flows of interest are those high enough to contact the channel banks (t hereby causing erosion and introducing MeHg to the water column), the water surface data coll ected in March is of pa rticular interest for surface water model calibration. However, the high flows at this time prevented the acquisition of thalweg elevations along the channel, and thus prevented an analysis of any relation between water surface slope and repres entative bed slope. For this reason, most channel bed data was collected during the pe riod of significantly lo wer flows. All data collected are associated with measurement e rrors. This error is re ported during the postprocessing stage. Unfortunately, some of the da ta points required multiple attempts at this 4503500 4503550 4503600 4503650 4503700 4503750 4503800 4503850 743800743900744000744100744200744300744400 Easting Coordinate (State Plane NAD 83, Nevada West)Northing Coordinate (State Plane NAD 83, Nevada West)Cross-Section 100 Cross-Section 200 Cross-Section 300 Cross-Section 400 Cross-Section 500 Flow Direction

PAGE 80

66 collection, as a result of errors outside of the acceptable range. These limitations, in addition to time constraints, resulted in a surveyed reach of basically 500 meters, including 5 complete transects. The spatial extent of the model domain is an important component to SMS model development. Surface Water Model Selection Based on cost considerations, research, a nd consultation with the model developers, it was decided that Surface Water Modeling Sy stem, Version 8.0 (SMS 8.0) would be the modeling system of choice for developing a su rface water model of the surveyed reach of the Carson River. Within the SMS model, se veral modules are available. The module of choice is FESWMS-2DH. The FE SWMS model executable code FLO2DH, is written in the Fortran 77 computing language. FLO2DH was originally developed by the USGS (Water Resources Division) and later modi fied by the Federal Highway Administration of the U.S. Department of Transportation. FESWMS-2DH applies the finite element meth od to solve the system of equations that describes 2D depth-averaged flow in a horizontal plane. The system can be used to simulate flow in rivers and floodplains a nd other surface-water bodies where vertical accelerations are small in comparison to horizontal motion (Froehlich, 2000). The hydrodynamic modeling code of FESWMS suppo rts both superand sub-critical flow analyses, including area wetting a nd drying. Use of this model is justified because one is not interested in the vertical variability of flow properties, but rather only the horizontal distribution of flow quantities. The temporal variability of flow within the Carson River channel also mandates the abil ity of the chosen model to account for wetting and drying patterns within grid elements along the ch annel boundary. Within this document, the

PAGE 81

67 surface water hydrodynamic mode l will be interchangeably referred to as SMS, FESWMS, or FLO2DH. Functionality of the FESWMS Model Several steps were involved in the de velopment of a functioning surface water hydrodynamic model. GPS data (Figure 5-1) wa s used to generate a quasi-3D finite element model of the reach of the Carson Ri ver immediately surrounding the Scout Camp and the groundwater wells. Channel bathymet ry data, from the field survey, was first used to generate a 2D finite element ne twork (or mesh) using the DIN2DH program. The finite element network was then refined through node addition a nd deletion, elevation modifications, and node spacing adjustments. With the model domain carefully planned, the final step was to simulate water flow through the model domain. Though the procedure is straightforward, the execution of this process for the Carson River model proved far more complicated. Several mont hs of research was placed into the development of this model, and each step of model development will be reviewed in turn. The primary equations governing the hydrodynamics are summarized by equations 5-1, 5-2, and 5-3 (Froehlich, 2000): Vertically-integrated momentum equation in the x direction: 0 ) ( ) ( 1 cos ) ( 2 1 ) cos (cos ) (2 2 y H x H HV x z gH HUV y gH HUU x t HUxy xx sx bx b x uv z x uu (Eq. 5-1) Vertically-integrated momentum equation in the y direction:

PAGE 82

68 0 ) ( ) ( 1 cos ) ( 2 1 ) cos (cos ) (2 2 y H x H HU y z gH HVU x gH HVV y t HVyy yx sy by b y uv z y vv (Eq. 5-2) Continuity equation q y HV x HU t H ) ( ) ( (Eq. 5-3) where variables are defined as follows ( units: [L] = length, [T] = time, [M] = mass): H = water depth [ L ] z b = bed elevation [ L ] z s = z b + H = water-surface elevation [ L ] U = depth-averaged velocity in th e horizontal x direction [ L / T ] V = depth-averaged velocity in th e horizontal y direction [ L / T ] uv, vu, vv = momentum flux correction coefficien ts that account for the variation of velocity in the vertical direction; coefficients result from the vertical integration of the momentum transport equations and account for the vertical variations of U and V x = arctan ( z b / x) y = arctan ( z b / y) z = arccos (1 – cos2x – cos2y) g = gravitational acceleration [ L / T2 ] = Coriolis parameter [ T-1 ] = water mass density, which is considered constant [ M / L3 ] bx, by = bed shear stresses acting in the x and y directions, respectively [ M / T2 L ] sx, sy = surface shear stresses acting in the x and y directions, respectively [ M / T2 L ] xx, yx, yy = shear stresses caused by turbulence where, for example, xy is the shear stress acting in the x directi on on a plane perpendicular to the y direction [ M / T2 L ] q = unit source (inflow) or unit sink (outflow) term [ L3/T / L2 ] These governing equations are solved by the FLO2DH program using the Galerkin finite element method applied to the user-defined grid network. Th e method is carried out using

PAGE 83

69 Newton-Raphson iteration and Gaussian qua drature integration for all elements (triangular or quadrangular shapes defined by nodes at vertices and mid-side points) within the model domain. FLO2DH is capable of simulating both steady and unsteady (time-dependent) 2D (in a horizontal plane) surface-water flow to obtain depth-averaged velocities and flow depths. The effects of be d friction and turbulen t stresses are included, as are, optionally, surf ace wind stresses and the Coriolis force (though the latter two are not used here). The postprocessor program AN O2DH is used to present the results of flow simulations either graphically or as re ports. Graphic plots of velocity and unit-flow vectors; ground-surface and water-surface level isolines; and velocity, unit flow, or stage (water-surface elevation) time-history plots at a computation point can be created. The principal gain from creation of this model is the generation of time-history plots of river stage at the lo cation of the groundwater well transect. An assessment can be made as to the agreement between river stag e over this reach as predicted by RIVMOD and FESWMS-2DH; this will be discussed in the comparative analysis portion of this document (chapter 8). Recall that the former (RIVMOD) is based on a generalized channel shape and data interpolated from topographic maps, while the latter (FESWMS2DH) is based on precise data collected usi ng GPS. The ultimate requirement for bank elements is to predict moisture content variab ility within the channe l banks, as a function of temporallyand spatially-v arying river stage, as pred icted by the current RIVMOD portion of IFMTFM. The FESWMS-2DH m odel was used to develop simple relationships between flow history and ba nk moisture, but the FE SWMS model is not directly incorporated into the IFMTFM; the simple relationships ar e rather applied to river stage as predicted by RIVMOD.

PAGE 84

70 Creation of Scatter Point Data Set Data points for a finite element mesh can be generated directly from topographic data, such as a list of survey points within a text file, imported into SMS. The Leica GPS software allows for the extraction of data point latitudes, longitudes, and elevations in the form of a simple text file. In this instance, the text file was a compilation of the survey data collected during March and July of 2001. Survey point data from the Leica output text files were first imported into Excel. Us ing examples of .xyz files included in the FESWMS documentation, the raw survey data was copied and manipulated within Excel to match the required input file format for FESWMS. Pre-Processing of Data Pr ior to Mesh Construction The FESWMS mesh was constructed by maki ng use of the available survey data. Before mesh creation, a moderate degree of pre-processing was required to account for shortcomings of the data set. Limited data downstream of cross-section 100 suggests that this cross section will serve as the downst ream boundary for the surface water model. Surveyed points downstream of this cross-s ection were therefore deleted. Erosion and deposition processes result in dynamic bed forms (high and low spots) and seasonal changes to exact thalweg location. In the in terest of consistency (and with the most complete data available in July), survey points along the channel bed collected during March (water surface elevations) were removed fr om the scatter data set. To preserve the valuable detail provided from the cross-sec tion surveys, however, the data points from the four downstream cross-sections collected in March are retained. D uplicate data points collected at nearly the same spatial location, as evidenced through visual inspection, were selected and deleted. While the software does include options for duplicate point removal, these means could not be used becau se of the spatial pr oximity of desirable

PAGE 85

71 points at the top and base of steep banks One is generally seeking a uniform nodal spacing in the range of 3 to 10 meters. Fine r resolution is needed in areas where the element slopes are steep, such as along the banks, and lower resolution is adequate for flatter regions, or where survey data is less extensive. The identification of points to be removed was tedious as each poi nt must be identified by co ordinates only, but trial and error indicated this to provide the best starting point for mesh generation, as these points determine the mesh boundary and the first cut for the mesh layout. With the point set modestly trimmed down, th e Excel file was saved as a text file, and the file extension was changed to .xyz. The survey points are now ab le to be read into FESWMS using the ‘File / Open’ command. Within the Fi le Import Wizard dialog box, the option was selected to open file as mesh and the coordinate systems are set to match the survey data (Horizontal System: Stat e Plane NAD 83, State Plane Zone: Nevada West – 2703, Vertical System: NAVD 88, Units: meters) before the file was opened. The data points from the file were converted to mesh nodes, points representing locations of known position and elevation. The key compone nts to a finite element mesh are the elements themselves, which connect the nodes to define the flow area, or model domain. The network of scatter data points (the dot s) and the associated mesh boundary is shown in Figure 5-2. For reference, the eleva tion contours created from this initial data set have also been overlain. All SMS-generate d figures are in color, with the scale and type of plotted contours (water surface el evation, water depth, ground elevation / bathymetry, or velocity) indi cated by the legend. A few items are of particular note. Vertical banks are poorly represented due to th e inability to properly survey the base of the banks, or the top of the banks, in certain instances. Evidence of the channel thalweg

PAGE 86

72 Figure 5-2. Original survey points after import into SMS scatter module (or low-flow inner channel) is present, but many of these points are disjointed. Additionally, the linear interp olation scheme of the model does not necessarily connect those points which should logically be connect ed. The plethora of points along the five channel cross-sections is also very clear, as is the paucity of points along some portions of the channel banks and bed. Clearly, nodal spacing is also far removed from uniform. The resolution of these shortcomi ngs is discussed at length in Appendix A Mesh Construction and Editing as Discussed in Appendix A The mesh construction and edit ing process is discussed in Appendix A (separate from the main body of text) not b ecause it was not a critical part of this thesis, but rather because the material is not needed for the r eader to understand and appreciate the results of this research. The material is availabl e for the reader’s reference for three main reasons: to provide proof that mesh constr uction was completed under a carefully planned set of guidelines, to provide explanation fo r the mesh’s limitation regarding the modeling elevation 1250.30 1250.80 1251.30 1251.80 1252.30 1252.80 1253.30 1253.80 1254.30 1254.80

PAGE 87

73 of overbank flows and the simulation of dynamic streamflows, and to justify the time spent getting the model to simulate Carson River hydrodynamics. The development of a functioning model based on this set of scattered survey data points proved to be far more labor-inte nsive task than initially believed. Model development required several months before the model was able to simulate streamflow conditions ranging from bank full levels to le vels at which flow was confined to the channel thalweg. Efforts were made to reta in channel morphology information contained within the survey data set. Attempts to model stage levels above bank full and on the floodplain proved unsuccessful. Several problem areas were encountered as next listed: model parameterization, including elemen t wetting and drying and specification of boundary conditions; steep channel banks, incl uding modified mesh boundaries and slope transitions; irregular channel morphology, as result from erosion and deposition processes; drying of elements internal to the wet / dry boundar y; jagged wet / dry boundaries; element re-wetting; supercritical flow conditions along boundary nodestrings; nodal velocity and water surface divergence; non-uniform node spacing and limited data availability for some portions of the mesh, resulting in poor element construction; and approaches for attaining model convergence. The mesh was constructed to accurately portray the geometry of the ch annel, as based on a combination of survey points, field measurements of channel bank dimensions, photographs, and personal recollection. The desired end-product was a model able to wet and dry as would be observed in the field. When the mesh ed iting process was complete, the FLO2DH hydrodynamic model was able to simulate element drying from bank-full to flows confined to the channel bed.

PAGE 88

74 Summary of Mesh Editing No amount of written explanation can expl ain the frequent frustration associated with getting this model to simulate the range of flows from bank-fu ll to nearly zero. Any change in mesh geometry, regardless of how tr ivial, required the model to be initialized with all elements submerged. Water surface lowering needed to be performed gradually to allow the time for model convergence. Iterat ion after iteration, attempts were put forth to identify as many potential problem areas as possible, but it was never known if the problem was corrected until the water surface wa s able to be lowered beyond a previous threshold. It remained uncertain if this wa s the last round of mesh editing or if many more divergence problems awaited discove ry. Eventually, however, the mesh was completed. The water surface was able to be lowered to minimum levels at which flow could no longer pass through the model domai n, due to interior elements along the thalweg becoming dry. After all the frustrati on, repetition, and uncertainty, the sense of satisfaction and accomplishment that is wo rthy of note. The final edited mesh (bathymetry) is shown in Figure 5-3. The well transect is indicated in the figure. For reference, the minimum flow levels able to be modeled are also shown in Figure 5-4. The flow magnitude at this condition is not discussed here as this early result was based on a 0.2-meter head gradient between the upstream and downstream boundary condition nodestrings, a gradient originally chosen fo r no other reason than to be conservative. With the model able to simulate element drying under these conditions, the model was predicted to be stable for a ny larger head gradient and the associated higher mean velocity. Admittedly, most of the required mesh editing was due to the manner in which the mesh was constructed, that being from a random network of survey points. While the model might have performed better had less deta il been included, this would also have

PAGE 89

75 significantly altered the reliability of the mode l results. After all, th e modeled reach of the Carson River is only 500 meters in length, a nd the goal was to accurately simulate the hydrodynamics (and thus river st age) down to low flows w ithin the channel thalweg, which necessarily requires a high degree of model detail. elevation 1250.17 1250.75 1251.32 1251.90 1252.47 1253.05 1253.63 1254.20 Figure 5-3. Bathymetry of Scout Camp r each after mesh editing found in Appendix A River Cross-Section at Well 1

PAGE 90

76 water depth : 130.000 0.01 0.37 0.72 1.08 1.43 1.79 2.15 2.50 Figure 5-4. Minimum flow levels able to be modeled, subject to a 0.2-meter head gradient

PAGE 91

77 CHAPTER 6 WATER TABLE RESPONSE TO STREAMFLOW VARIABILITY Linking Surface Water Modeling and Groundwater Data When the SMS model was selected as the surface water model for this project, two distinct end-products were envisioned. The firs t was to develop a simple relationship to relate stage in the channel to water table level within the bank. This relationship was then to be incorporated into a new IFMTFM bank element functionality to predict saturation conditions within the channel bank immediately adjacent to the channe l, as a function of flow depth and channel geometry known by RIVMOD. The second end-product, to be discussed in chapter 8, was a comparative analysis between RIVMOD predictions and SMS predictions. For the modeled Scout Camp reach, a comparison between the two models was sought for flow depths and mean velocity to see if the simplified channel geometry and 1D nature of RIVMOD results in consistently hi gh or low predictions relative to a more detailed and calibrated model. Data Availability A GPS survey was conducted to provide cha nnel bathymetry input data to the SMS model, which was used to model Carson River streamflows and predic t river stage. After site selection in March 2001, the survey operation began. W ith the groundwater wells sited, the intention was to survey channel cross-sections both upstream and downstream of the well transect, with roughly equivale nt survey areas in both directions. Complications arose with loss of satellite lock caused by cottonwood trees, vertical banks, and, at times, less-than-i deal satellite configurations Further complications arose

PAGE 92

78 when post-processing indicated points with un acceptable error values, which would have to be surveyed again. After two channel survey operations in March 2001 and July 2001, five channel cross-sections and approxim ately 500 meters of channel length were surveyed. Three groundwater wells are situ ated along a transect, approximately 275 meters downstream of the upstream boundary condition of the mesh, as indicated in Figure 5-3. The SMS mesh was constructed ( Appendix A ) to predict the relationship between streamflow and river stage, sp ecifically the stage adjacent to the vertical bank at the well transect. Three gages along the Carson Ri ver monitor streamflow levels. The RIVMOD module in IFMTFM incorporates dynamic stre amflow values from the Carson City and Fort Churchill gages to perform hydraulic ro uting. Upstream irrigation withdrawals are accounted for using the Fort Churchill str eamflow values. With no known withdrawals downstream of Fort Churchill, RIVMOD is capable of routing flows and predicting average streamflows for all RIVMOD segmen ts downstream of Fort Churchill. Generation of Boundary Condition for FESWMS Surface Water Model To develop a relationship between river stag e and water table level, the river stage needed to be accurately modeled within th e FESWMS module of SMS. Following mesh creation, editing, and refinement; but befo re simulating the hydrodynamics through the mesh reach, boundary conditions must be specified. The FLO2DH model requires an upstream and a downstream boundary condition. During the mesh construction process ( Appendix A ), it became evident that the finite-element mesh was not suitable for the simulation of dynamically rising and falling flows. The decision was made to use head boundary conditions at both the upstream and downstream extents of the model spatial domain. Nodestrings al ong the boundaries and

PAGE 93

79 strategically located at crosssections within the mesh inte rior were used along with the model solutions for flow depth to identif y the flow through any given cross-section. Water surface elevations from a FESWMS model solution are, therefore, relatable to streamflow through the model domain. Within FESWMS, the head boundary conditions are incrementally lowered to generate a series of water surface soluti ons. The FESWMS module can stably operate with a head or flow upstream boundary c ondition, but the model requires a head downstream boundary condition. Because the Scout Camp site is not located at a gaged location, a rating curve at the downstream boundary was not available. The head gradient between upstream and downstream boundary conditions must, instead, be specified and held at a constant value, based on the lack of stage data for the downstream boundary and a lack of data for the variability of wate r surface slope with streamflow magnitude, as later discussed. Though an upstream head boundary is used in the surface-water model, the flow hydrograph at the upstream boundary ch annel cross-section must also be known in order to later link surface water modeling of river stag e to the corresponding water table elevations from the we lls within the channel bank. Streamflow routing: testing the valid ity of instantaneous translation Water table data collected via the groundwat er wells is a time series of water table elevation (NAVD 88). Flow in the channel also varies over time, as evidenced by the hydrographs of USGS gaging stations at Carson City and Fort Churchill. This flow must be specified to develop a corre lation between water table and river stage. Two options to obtain this time series were considered. Firs t is the possibility of an evapotranspiration (ET) forcing function. Under this scenario, st reamflow would simulta neously rise and fall at all locations along the Carson River. Peak flow would occur everywhere at the same

PAGE 94

80 time of day. Differences in streamflow magnit ude with distance are th e result of irrigation withdrawals and tributary addi tions. Given the arid envir onment and the prevalence of transpiring cottonwood trees along the entire river length, this seemed a viable possibility. Streamflow withdrawals for i rrigation are believed to be nonexistent downstream of the Fort Church ill gaging station. If the ET forcing function was indeed present, so that streamflow oscillation is solely due to flow withdrawals, then the streamflow time series at Fort Churchill could be instantaneously translated (no consideration of travel time ) and used as the upstream boundary condition for the Scout Camp reach. This instantaneous downstream translation by a distan ce of 10 km may not seem intuitively correct, as a reasonable velocity of 0.5 m/s would translate to a 6-hour travel time. However, it would simplify the effo rt to obtain the streamflow time series at the upstream boundary, and the option was investigated. To validate or invalidate the assumption of instantaneous translation and an ET forcing function, streamflow data from thr ee Carson River gages were compared for the period of available well data (March 2001 through September 2002). The three USGS gages, Carson City Gage (CCG), Deer Run Road Gage (DRR), and Fort Churchill Gage (FCH), record streamflow at 15-minute intervals. This set of data was obtained from the USGS with the assistance of USGS Hydrol ogist Jim Crompton in Carson City, Nevada. The streamflow time series are graphed in Figure 6-1 (2001) and Figure 6-2 (2002). The data is separated by calendar year. A logarithmi c scale is used to better capture the high degree of streamflow variability over this peri od of interest. Streamflow data available at 15-minute resolution should capture the diel (with in day) variability at these respective locations.

PAGE 95

81 0.1 1 10 100 1000 10000 3/7/20014/6/20015/6/20016/5/20017/5/20018/4/20019/3/200110/3/200111/2/200112/2/20011/1/2002 date and timeCarson River Streamflow (cfs) Carson City Gage Fort Churchill Gage Deer Run Road Gage Figure 6-1. 2001 USGS streamflow ga ge comparison: 15-minute data Figure 6-2. 2002 USGS streamflow ga ge comparison: 15-minute data 0.1 1 10 100 1000 10000 1/1/20021/31/20023/2/20024/1/20025/1/20025/31/20026/30/20027/30/20028/29/20029/28/2002 date and timeCarson River Streamflow (cfs) Fort Churchill Gage Carson City Gage Deer Run Road Gage

PAGE 96

82 0 200 400 600 800 1000 1200 1400 1600 5/1/20025/4/20025/7/20025/10/20025/13/20025 /16/20025/19/20025/22/20025/25/20025/28/20025/31/2002 date and timeCarson River Streamflow (cfs) Carson City Gage Deer Run Road Gage Fort Churchill Gage Figure 6-3. May 2002 Carson River streamflow at USGS gages The complete set of data provides evidence of a travel time for flow peaks from upstream gaging stations to Fort Churchill. A more noticeable time lag is evident for DRR to FCH as compared to CCG to DRR, as reasonably expected given the 6-km versus 52-km distance between the respective stations. A subset of the data, for May 2002, is used to illustrate this point, appeari ng in Figure 6-3. Note the diel variations in streamflow. The daily peaks do not occur consistently at the sa me time of day, but rather occur at times dictated by travel time from upstream. During times of year when streamflow is primarily due to snowmelt, higher melt rates would be observed during the early afternoon when the day is at its hottest As the melt rates slow overnight, the result is a diel streamflow variability which w ill then characterize the observed streamflow further downstream. Despite the arid climate, the diel streamflow variability caused by ET is clearly overshadowed by the influence of travel time. The observed streamflow

PAGE 97

83 reduction from CCG to FCH is due to irriga tion withdrawal; no know n irrigation ditches or tributaries occur from Fort Churchill to the study reach at the Scout Camp. The Scout Camp site is approximately 10 km downstream from FCH, so a lag time is expected between these two locations. Fort Churchill streamflows will need to be hydraulically routed to create the upstream boundary condi tion for the SMS model. The assumption of instantaneous translation would have the effect of artificia lly impacting the observed time lag between modeled river stage (a function of streamflow) and water table oscillations. In the extreme case of rapidly rising flow and rapid response time between channel and bank, the assumption of instan taneous translation could presumably cause the well response to precede the river stage oscill ation. This approach was discarded. Streamflow routing: hydraulic routing using RIVMOD Streamflow was instead routed usin g RIVMOD. RIVMOD pe rforms hydrodynamic computations using a 60-second time step. Being only interested in the hydrodynamic output from IFMTFM, there was no need to ac curately simulate reservoir stage, which impacts the mercury computations. This alleviated the need to obtain and process 15-minute data for the USGS gage in Lahontan Reservoir at the dam (LRD). A time series for flow must still be input to RIVMOD at LRD, but the values were simply set equal to the time series for flow at FCH. This approach ensures that backwater conditions from the reservoir do not impact the hydrau lic routing from Fort Churchill to the reservoir, between which the Scout Camp is located. Within IFMTFM, streamflow values input to RIVMOD are average daily flows. To be consistent with the 1-hour interval at wh ich water table data is collected by the well data loggers, a 1-hour time interval was se lected for streamflow input to RIVMOD for the hydrodynamic routing. To create the RIVMOD input data files, macros were written

PAGE 98

84 to extract hourly flow data from the sma ller 15-minute data sets. Data from CCG and FCH were extracted, and, as mentioned, flows at LRD are taken as equa l to those at FCH. Inflow to the reservoir from the Truckee Ca nal was modeled as zero. Care was taken to ensure that the flow values at the respectiv e stations within any given input file row corresponded to the same point in time, as n eeded to compensate for the intermittent data gaps within the 15-minute data files. Flows lo wer than 8.0 cfs were changed to 8.0 cfs, as this value represents the minimum allowa ble flow input for RIVMOD within IFMTFM. Periods of lower flow are therefore no t accurately simulated by RIVMOD. Before simulating real-time hourly streamflow da ta, RIVMOD was run at steady state for a period of 10 days to create a valid in itial solution throughout the RIVMOD model domain from CCG to Lahontan Reservoir. Hydrodynamics were then simulated for the period from March 2001 through August 2002 usi ng hourly gage data, and the solution time series for flow are output, to a text file, for RIVMOD segments 148 through 155 (segments near Scout Camp reach). It is also noted that two separate hydrodynamic simulations were necessary because of file size limitations in RIVMOD; a 30 day overlap was used to ensure flow continuity when the simulation results were later combined. Latitude and longitude of all survey points are known from the survey data set in Nevada West State Plane Coordinates. RIVMOD segments have previously been recorded on a USGS topographic map, whic h utilizes the 1927 No rth American Datum (NAD 27) as the horizontal reference. Within SMS, the mesh coordinates (surveyed date points) are easily converted to an alternate coordinate system from the Edit menu. This feature was used to identify the NAD 27 c oordinates corresponding to the upstream and downstream boundary conditions, as well as the location of the well transect. The

PAGE 99

85 upstream and downstream boundary conditions are located within RIVMOD segments 151 and 153, respectively. The groundwater well transect is approximately centrally located within RIVMOD segment 152. Segm ent spacing for model segments 144 through 240 is equal to 250 meters. RIVMOD flow computations we re used to construct the flow time series at the upstream boundary condition of the FESWMS m odel. With no significant inflows or flow withdrawals downstream of Fort Chur chill, hydraulic routing results in hydrograph attenuation and delay due to channel resistance. RI VMOD does not model hydrograph attenuation due to storage capacity within channel banks. RIVMOD is able to smooth the FCH hydrograph, which can oscillate significan tly during periods of rapidly changing streamflow or under low flow conditions (b ecause of inherent gage limitations). The hydrograph input into RIVMOD uses hourly ti me steps. The routed hydrograph output from RIVMOD is extracted at hourly inte rvals, but all RIVMOD computations are completed at 60-second time steps. Groundwater Well Data General Well Information Three groundwater monitoring wells were installed in March 2001, approximately centrally located within the surveyed Scout Camp reach. The 6-inch diameter, 20-foot deep wells are located along the south streamba nk at distances from the vertical bank of approximately 5 (Well 1), 25 (Well 2), and 100 (W ell 3) feet. The three wells are situated along a transect perpendicular to streamflow direction. Pressu re transducers situated at the base of each well monitor the water table level, with data con tinuously recorded by data loggers. With th e notable exception of the middle we ll, continuous water table data is generally available dating back to the time of we ll installation.

PAGE 100

86 Data Limitations Although well installation was completed on March 9, 2001, time was allowed for the water to clarify prior to data logger insertion. Collecti on of water table data began on March 12, 2001, after the March 2001 field su rvey work was completed. Water table elevations, relative to the NAVD 88 vertical datum, are meas ured in each well at hourly intervals. The probe installed in Well 2 failed to function properly, and this data set is therefore unavailable. After the failure of th is probe, the remaining two probes were also removed on May 29, 2001 for preventative mainte nance, to ensure no similar problems in the future. The probes were unable to be retu rned to the wells before the field work period in July 2001. For this reason, a data gap appears in the water table data sets for Well 1 and Well 3 between May 29 and July 5, 2001. During the reinstallation process, the cable for holding up the probe in Well 1 was insufficiently stabilized. Duct tape failed to hold the looped cable in place, and the probe weigh resulted in the probe falling to the base of the well. As this process occurred gr adually, a uniform correct ion is unable to be applied to the well data. Probe s lippage is evident in the well data for this time period, as shown in Figure 6-4, and was verified by well inspection in October 2002 (note the sudden rise in water table, inconsistent with water table oscillations observed in Well 3, associated with the increased water pressure felt by the transducer as it moved down the well). With the probe slippage complete and the probe at the base of Well 1, the slip distance was calculated and app lied as a uniform correction to the water table data for Well 1 after August 1, 2001. The 2001-2002 water ta ble elevations in Figure 6-4 are plotted relative to the thalweg elevation at the well transect (1252.01 meters (NAVD 88) set as point of zero-reference); additionally in cluded for reader reference are the top of bank elevation (1254.15 meters), the top of casing elevation for Well 1 (1253.95 meters),

PAGE 101

87 and the high point of the channel bed along the well transect (1252.27 meters), which is indicative of the elevation at which the cha nnel bank closest to Well 1 will no longer be in contact with water actively fl owing through the main channel. -2.000 -1.500 -1.000 -0.500 0.000 0.500 1.000 1.500 2.000 2.500 3/1/20014/30/20016/29/20018/28/200110/27/200112/26/ 20012/24/20024/25/20026/24/20028/23/200210/22/2002 date and timeElevation Relative to Thalweg (meters) Well 1 (5' from bank) Well 1 Error Well 3 (100' from bank) Top of Bank Transect Thalweg Probe Slippage, Data Points Invalid Figure 6-4. Complete water table data from Well 1 and Well 3 at Scout Camp site Concern with computing the hydraulic gr adient between the river and the bank sediments directed the selection of the useable set of well data as data is only selected from time periods in which water table data is available in both Well 1 and Well 3. This data set is used in the later analysis of th e coupling of river stage and water table level. Complete water table data is available for the follow ing time periods: March 12, 2001 through May 9, 2001 and August 1, 2001 through Oc tober 22, 2002, the date of the final site visit. The loss of data from May 9, 2001 through July 5, 2001 was unavoidable, and the loss of data from July 5, 2001 through August 1, 2001 was for a period of low streamflow, below the point of contact with the banks. Recall, the relationship between

PAGE 102

88 river stage and saturated bank levels is of particular interest, so the latter data loss is considered trivial. Fifteen-minute streamflow data, obtained for the period from March 1, 2001 through August 29, 2002, was used to creat e RIVMOD input data files for hydraulic routing and to develop a hydrograph at the upstream boundary of the SMS mesh. Assimilation of this data took place in Se ptember of 2002. Given that flows after August 29, 2002 are very low and below the point of bank contact, no effort was deemed necessary to update the 15-minute data set fo llowing the download of additional well data in October 2002. Water table response to st reamflow oscillations after August 29, 2002 will not be considered in the data analysis po rtion of this project. The complete set of overlapping well data for Well 1 and Well 3 is shown in Figures 6-5. Note that the water table elevations appear rela tive to the NAVD 88 vertical da tum (not thalweg elevation). The thalweg and bed high point elevations along the well tr ansect are again shown for reference. The graphs do evidence a travel tim e, related to hydraulic conductivity of bank sediments and signal dampening with in creasing distance from the channel. Observed Hydraulic Gradient within Bank Sediments Before any data was collected, it was assu med that the river stage would influence the water table levels within the channel bank, as observed at the groundwater wells. This direct response relationship re quires the hydraulic gradient to be in the direction away from the channel. The hydraulic gradient be tween Well 1 and Well 3 is computed as the difference in the respective water table elevat ions for each hourly sample interval, such that a positive value indicates a gradient aw ay from the channel. As shown in Figure 6-6, the hydraulic gradient is indeed away from the channel, with the singular exception of October 23, 2001 through November 19, 2001, during which the value is barely negative. During this time period, the measured flow at FCH is less than 2.0 cfs. This low flow

PAGE 103

89 1250.000 1250.500 1251.000 1251.500 1252.000 1252.500 1253.0003/1/20014/30/20016/29/200 18/28/200110/27/200112/26/20012/24/200 24/25/20026/24/20028/ 23/200210/22/2002date and timeElevation, NAVD 88 (meters) Well 1 Well 3 Bed High Point Transect Thalweg Figure 6-5. Transient water le vels in Well 1 and Well 3, 2001-2002 cannot be routed using RIVMOD, but it is likely that the rive r is no longer flowing at the Scout Camp. Any water in the channel is likely the result of baseflow from the connected surficial aquifer, which may explain the brief gradient towards the river. For all periods for which the channel banks are contacted (perio ds of interest), the hydraulic gradient is away from the river. This verifies the or iginal assumption that saturation conditions within the channel banks are determined in response to changes in river stage. Pump Test Analysis Additional field data for the analysis of water table response to streamflow oscillations was obtained from a pump test conducted at the Scout Camp site in October 2001. A submersible pump was lowered to near the bottom of Well 1. Pressure transducers with data loggers were used in Well 1 and Well 3 to record the water table

PAGE 104

90 Figure 6-6. Hydraulic differential from Well 1 to Well 3, 2001-2002 response within the wells during the pump test The data logger in Well 1 was attached to a laptop so that water levels could be mon itored throughout pump test duration. Flow rate was adjusted to ensure the drawdown level would not fall below the pump intake, while also allowing the drawdown to approach st eady state within a reasonable time duration. Prior to recording pump test da ta, the pumping flow rate was allowed to stabilize, given that a constant flow rate is necessary for data analysis. The drawdown versus time relationship obser ved for Well 1 during the pump test is shown in Figure 6-7. The pump test was c onducted for a duration of 112 minutes at a flow rate of 1 gallon per 32 seconds (0.00418 cf s), sufficient time for the water table to approach steady state. Generally, three approa ches can be followed in analyzing the response of water levels in an unconfined a quifer during a pump test (Fetter, 2001). The -0.250 0.000 0.250 0.500 0.7503/1/20014/30/20016/29/20018/28/200110/27/200112/26/20012/24/20024/25/20026/24/20028/23/200210/22/2002date and timeWater Elevation Difference (meters)

PAGE 105

91 0.0 0.5 1.0 1.5 2.0 2.5 0.010.11101001000 elapsed time (minutes)observed drawdown (meters) Figure 6-7. Pump test drawdown data collected for Well 1 in October 2002 first approach uses a 2Dor 3D-variably saturated flow model, but the detailed soil characteristic data needed for this type of analysis is not available for this study. The second approach requires the assumption of small drawdown for an observation well some specified distance away from the pum ped well. To validate this assumption, two pieces of data are necessary: observed drawdown versus time data for an observation well and knowledge of the saturated thickness of the unconfined aquifer. For the Scout Camp field site along the Carson River, no draw down was observed at Well 3 (likely the result of the small pumping rate), at a distance of 95 feet from Well 1; and the well penetration is only 25 feet, with the total saturated thickness unknown. The third option, as will be used in this analysis, bases calculations on a match point between the time-drawdown data and theoretical curves.

PAGE 106

92 The flow of water in an unconfined aqui fer toward a pumping well is described by equation 6-1 (Neuman and Witherspoon, 1969): (Eq. 6-1) where h is the saturated thickness of the aquife r [L], r is the radial distance from the pumping well [L], z is the elevation a bove the base of the aquifer [L], Ss is the specific storage [1/L], Kr is the radial hydraulic conductivity [L/T], Kv is the vertical hydraulic conductivity [L/T], and t is time [T]. Neuman (1975) has published a solution to the above equation. The solution incl uded two parts, one for the time just after pumping has begun (water is coming from specific storage) and one for much later (when the water is coming from gravity drainage of aquifer por es) (Fetter, 2001). Neuman’s solution is written as equation 6-2: ) , ( 40 B Au u W T Q h h (Eq. 6-2) where (W(uA,uB, )) is the well function for the wa ter table aquifer, based on the dimensionless parameters defined by equations 6-3, 6-4, and 6-5: Tt S r uA42 (early time drawdown data) (Eq. 6-3) Tt S r uy B42 (late time drawdown data) (Eq. 6-4) h vK b K r2 2 (Eq. 6-5) where (h0 – h) is the drawdown [L], Q is the pumping rate [L3/T], T is the aquifer transmissivity [L2/T], S is the storativity [dimensionless], Sy is the specific yield t h S z h K r h r K r h Ks v r r 2 2 2 2

PAGE 107

93 [dimensionless], b is the initial saturated thickness of the aquifer [L], and other parameters are as previously defined. After plotting pump test dr awdown versus time using a lo g-log scale, the data was overlaid on the theoretical curves pres ented by Neuman, and a match point was determined. The following parameter values de scribe the selected match point for the early time drawdown data: W(uA,uB, ) = 1.80, 1/uA = 5.5E10-5, = 0.017, t = 0.10 minutes, and (h0 – h) = 1.0 meters. Utili zing the presented equations and these parameter values, with the pumping rate, it is possibl e to solve for aquifer transmissivity: day m u W h h Q TA 2 046 1 ) ( ) ( 4 (Eq. 6-6) Obtaining an estimate for hydraulic c onductivity requires kn owledge of the saturated aquifer thickness. With field meas urements of the water table immediately before pumping and the elevation of the base of the pumped well, a minimum saturated thickness is calculated as 5.2 meters. A low-end estimate for horizontal hydraulic conductivity is then calculated: s cm day m b T Kimum h 4 min10 25 3 28 0 (Eq. 6-7) Principal aquifers in the Carson River Ba sin are unconsolidated basin-fill deposits of interbedded gravel, sand, silt, and clay (Covay et al., 1996). The above estimate for horizontal hydraulic conductivity is a spat ial average over the portion of the aquifer impacted by the pump test. Not surprisingly, the calculated value falls within typical ranges for the following sediment types: si lt, sandy, silt, clayey sands, till (10-6 to 10-4

PAGE 108

94 cm/s); and silty sands, fine sands (10-5 to 10-3 cm/s) (sediments characteristic of typical alluvium) (Fetter, 2001; Freeze and Cherry, 1979). Channel bank sediments are heterogeneous throughout the Carson River system, a point verified by inspection of bank soils along the Scout Camp site. Recall (chapter 1) that the Carson River underwent significant changes in both channel form and process during the period of the Comstock Lode, resu lting in the storage of mercury-enriched sediments within a complexly structured all uvial sequence located all along the Carson River valley downstream of Carson City. Today, the Carson River continues to experience cycles of sediment erosion and de position associated with high flow events, processes which continue to pr oduce sediment heterogeneity. The pump test estimate of average hydraulic conductivity is not used as a determining factor for water table response to streamflow variability. With detailed sediment cores unavailable for the field site, acceptance of this number as representative is difficult. It is conceivabl e that the pump test is onl y extracting water through a high hydraulic conductivity layer, such as a deposited layer of coarse sand or a sediment-filled meander of a historic channel within the hyporheic z one, rather than from sediments truly representative of the surficial aquifer. The calculated hydraulic conductivity from the pump test may not be representa tive of the surficial aquifer, a point further reinforced by the rapid water table recovery to zero dr awdown within 15 minutes of the cessation of pumping, as shown by the late time data point s in Figure 6-7. The pump test analysis was used only in a qualitative sense to gain insigh t to the system. A more useful analysis is provided from a direct comparison of routed streamflow and well data for the Scout Camp reach of the Carson River.

PAGE 109

95 Quantifying the Relationship between River Stage and Water Table Elevation It has been suggested that streamflow osci llations, with the associated changes in river stage, will result in an observed respons e in the water table w ithin the channel bank. The water table level is not, admittedly, a perfect indicator of saturation conditions within the channel banks. The hydraulic conductivity of bank sediments will determine the rate at which the bank becomes saturated when river stage is rising and the rate at which the bank dries as river stage fa lls. Knowing that the SMS model would be unable to dynamically simulate river stage rising and falling, and given that the mesh remained incomplete, alternative means to analyze the degree of conn ectivity between channel and bank water levels began to be considered. Although the relationship between streamflow and river stage at the well cross-section coul d not yet be predicted using the SMS model, there was obviously a direct relationship between streamflow and river stage. Comparison of Streamflow Hydrograph and Water Table Response The routed streamflow h ydrograph from RIVMOD is the only data available for streamflow in the vicinity of the groundwat er wells within the surveyed Scout Camp reach. The streamflow values are not perf ect substitutes for field measurements of streamflow because of the us e of a 1D model, simple input channel geometry, no bank storage considerations, and sp atial averaging over 250-mete r segment lengths; but this data should be a reasonable representation of the routed hydrogra phs. Visual inspection of a portion of the routed RIVMOD hydrograph shows evidence of travel time and modest magnitude attenuation from the Fort Churchill gage to the Scout Camp reach. Hydrograph routing from Fort Churchill to th e Scout Camp for the streamflows of 2001 and 2002 are shown in Figures 6-8 and 6-9, re spectively. To show enhanced detail, hydrograph routing for streamflows of Apr il 2002 are shown in Figure 6-10. RIVMOD

PAGE 110

96 Figure 6-8. 2001 streamflow routing co mparison: FCH gage data and RIVMOD simulation for upstream boundary condition segment 150 is selected, as it is the first segment upstr eam of the modeled upstream boundary for the SMS mesh. Travel time from Fort Churchill to the Scout Camp, a distance of 10 km, ranged from 2 to 8 hours for the observed 2001-2002 streamflow magnitude range from 0 to 1500 cfs. The ups tream boundary is situated in segment 151, the groundwater wells in segment 152, and the downstream boundary in segment 153. The hydrograph for segment 150 was selected fo r one additional reason. As it turns out, routed flows for segments 151 or 152 would pr edict well response to occur in advance of the associated streamflow change in a few in stances. Given the simplif ications inherent to RIVMOD (most notably no inclusion of hydr ograph attenuation due to bank storage effects), noting that the RIVMOD segments ar e only 250 meters in length (indicative of a small travel time from one segment to the next), realizing that well data is collected only at 1-hour time intervals (leaving uncertainty between any consecutive data points), and noting that Well 1 is 5 feet removed from the vertical channel bank, this could certainly 0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 900.00 1000.00 1100.00 1200.00 1300.00 1400.00 1500.003/1/20013/31/20014/30/20015/30/20016/29/20017/29/20018/28/20019/27/200110/27/200111/26/200112/26/2001Date and TimeRouted Flow (cfs) Fort Churchill Segment 150

PAGE 111

97 Figure 6-9. 2002 streamflow routing co mparison: FCH gage data and RIVMOD simulation for upstream boundary condition 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 4/1/20024/4/20024/7/20024/10/20024/13/20024/16/ 20024/19/20024/22/20024/25/20024/28/20025/1/2002 Date and TimeStreamflow (cfs) Fort Churchill Segment 150 Figure 6-10. April 2002 streamflow routing comparison: FCH gage data and RIVMOD simulation for upstream boundary condition 0.00 100.00 200.00 300.00 400.00 500.00 600.00 700.00 800.00 900.00 1000.00 1100.00 1200.00 1300.00 1400.00 1500.001/1/20021/31/20023/2/20024/1/20025/1/20025/31/20026/30/20027/30/20028/29/2002Date and TimeRouted Flow (cfs) Fort Churchill Segment 150

PAGE 112

98 have been predicted as a likely possibility. It has already been s hown that the hydraulic gradient is away from the channel, so that water table levels will ch ange in response to a rising or falling river stage. Stage change must therefore precede water table response. The near-channel (Well 1) response to streamflow variability during calendar years 2001 and 2002 is shown in Figures 611 and 6-12. The routed hydrograph from RIVMOD segment 150 is indicated as th e Scout Camp UBC (upstream boundary condition) in these and the following plots. To view the degree of connectivity between streamflow and Well 1 water levels, it is neces sary to zoom in on these annual plots. A complete set of figures is found in Appendix B in temporal order from March 2001 through June 2002. This collection of figures is used to show the following: RIVMOD hydrograph routing from Fort Churchill to the Scout Camp reach for each month of simultaneous record; Well 1 water table res ponse to streamflow variability for each month of simultaneous record; and Well 1 water table response to streamflow variability for selected periods of intere st on the order of a few days. These periods of interest are selected based on two characteristics: high degree of streamflow oscillation; and distinct peaks and troughs. Months excluded from the appendix are those in which streamflows are low in magnitude (such that banks are no longer contacted) a nd connectivity is less evident. Note the dual y-axes showing both Carson River streamflow and water table elevation; scales were adjusted as need ed to best indicate the time lag between streamflow change and water table response. A sampling of these plots is shown in Figures 6-13 through 6-23. The selected figures demonstrate the water table response for streamflow variability for tw o specific flow classifications : flood events (example: April to May 2002, Figures 6-13 through 6-21), where flows are highly dynamic, and

PAGE 113

99 periods of nearly constant flow (example: February 2002, Figures 6-22 and 6-23), where fluctuations are diurnal in nature and likely due to ET. Indicated by this small sampling and the complete set of figures in Appendix B is the fact that peaks and troughs for routed st reamflow and well data occur at essentially the same time. This time will not be exactly the same, but the time of peak or trough will occur during the same hour because RIVMOD a nd well data are both available hourly at times which do not exactly overlap. Streamflow changes in all instances precede well response. No evidence of a time lag from stag e change to water ta ble change (at least within the data frequency lim itation of 1 hour), is found with in the data. This observation does not state that the water le vel within the channel and with in the bank will be the same (in fact, it cannot be the same given the 5foot separation from channel to well and evidence of a hydraulic gradient away from the channel). 0 200 400 600 800 1000 1200 1400 1600 1800 2000 3/7/20014/6/20015/6/20016/5/20017/5/20018/4/ 20019/3/200110/3/200111/2/200112/2/20011/1/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elevation, NAVD 88 (meters) Scout Camp UBC Streamflow of 173.8 cfs Well 1 Water Level Thalweg at Well Cross-Section Figure 6-11. 2001: near-channel well re sponse to streamflow variability

PAGE 114

100 0 200 400 600 800 1000 1200 1400 1600 1800 2000 1/1/20021/31/20023/2/20024/1/20025/1/20025/31/20026/30/20027/30/20028/29/20029/28/200210/28/2002 date and timeCarson River Streamflow (cfs)1250 1250.5 1251 1251.5 1252 1252.5 1253 1253.5Water Table Elevation, NAVD 88 (meters) Scout Camp UBC Streamflow of 173.8 cfs Well 1 Water Level Thalweg at Well Cross-Section Figure 6-12. 2002: near-channel resp onse to streamflow variability Previously, there was thought of a time lag between river stage change and water table response, and the thought further stated that this ti me lag could vary depending on whether flows were rising or falling. For example, if the soil is wetting (river stage rising), clays in the bank sediments could sw ell and slow infiltration of water into the channel bank. After extended periods of lo w flow and dry conditions, surface crusting could impede water infiltration into the banks Rapidly-rising streamflows could entrap air within the banks and effectively redu ce the hydraulic conductivity of the bank sediments. Rapidly falling flows could show a lag time prior to water table dropping due to bank storage phenomena. With no visual ev idence of a time lag for flow peaks or troughs, well response is clearly tightly coupl ed to streamflow va riability. This high degree of coupling holds true dur ing flood events, periods of n early constant flow, and

PAGE 115

101 0 200 400 600 800 1000 1200 1400 1600 1800 2000 4/1/20024/4/20024/7/20024/10/20024/13/20024/16/ 20024/19/20024/22/20024/25/20024/28/20025/1/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-13. April 2002: n ear-channel response to streamflow variability 0 100 200 300 400 500 600 4/1/20024/2/20024/3/20024/4/20024/5/20024/6/20024/7/20024/8/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-14. April 1-7, 2002: near-channe l response to streamflow variability

PAGE 116

102 Figure 6-15. April 8-14, 2002: near-channe l response to stream flow variability 0 200 400 600 800 1000 1200 1400 1600 1800 4/15/20024/16/20024/17/20024/18/2002 date and timeCarson River Streamflow (cfs)1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8 1252.9Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-16. April 15-17, 2002: near-chann el response to streamflow variability 0 100 200 300 400 500 600 700 800 900 1000 4/8/20024/9/20024/10/20024/11/20024/12/20024/13/20024/14/20024/15/2002 date and timeCarson River Streamflow (cfs)1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level

PAGE 117

103 0 200 400 600 800 1000 1200 1400 1600 1800 2000 5/1/20025/4/20025/7/20025/10/2002 5/13/20025/16/20025/19/20025/22/20025/25/20025/28/20025/31/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-17. May 2002: near-channel re sponse to streamfl ow variability 0 100 200 300 400 500 600 5/1/20025/2/20025/3/20025/4/20025/5/20025/6/20025/7/20025/8/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-18. May 1-7, 2002: near-channe l response to streamflow variability

PAGE 118

104 0 100 200 300 400 500 600 700 800 5/8/20025/9/20025/10/20025/11/20025/12/20025/13/20025/14/20025/15/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-19. May 8-14, 2002: near-channel response to streamflow variability 0 200 400 600 800 1000 1200 1400 1600 1800 5/15/20025/16/20025/17/20025/18/20025/19/20025/20/20025/21/20025/22/2002 date and timeCarson River Streamflow (cfs)1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-20. May 15-21, 2002: near-channe l response to stream flow variability

PAGE 119

105 0 200 400 600 800 1000 1200 1400 1600 1800 5/25/20025/26/20025/27/20025/28/20025/29/20025/30/20025/31/20026/1/2002 date and timeCarson River Streamflow (cfs)1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-21. May 25-31, 2002: near-channe l response to stream flow variabililty 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2/1/20022/4/20022/7/20022/10/20022/13/20022/16/20 022/19/20022/22/20022/25/20022/28/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure 6-22. February 2002: near-channe l response to streamflow variability

PAGE 120

106 0 50 100 150 200 250 2/1/20022/2/20022/3/20022/4/2002 date and timeCarson River Streamflow (cfs)1251.6 1251.7 1251.8 1251.9 1252.0 1252.1 1252.2 1252.3Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure 6-23. February 1-3, 2002: near-cha nnel response to streamflow variability for intermediate streamflows during the peri od of record (2001-2002) at the Scout Camp site. RIVMOD is not the perfect tool for hydraulic routing becau se of the implicit simplifications, and the well data is not pe rfect because of the 1-hour data collection intervals, but, regard less, there is no evidence for a tim e lag on the order of hours (with 1 hour as the lower bound based on da ta collection frequency). If this hydrograph for RIVMOD segment 150 were applied as the upstream boundary condition in the SMS model, a slight time lag would necessarily be added within SMS to translate the upstream boundary hydrograph to the well cross-section. This would further reduce the time lag shown in the above figures and Appendix B This comparison is not perfect because RIVMOD r outing is 1D and incorporates coarse channel geometry. The hydrograph from a m odel segment 500 meters upstream of the well cross-section is also being used for the comparison with the well data. The time lag

PAGE 121

107 cannot be negative because the hydraulic gradient is away from the channel, and segment 150 is the first segment upstream of the well ensuring this to be true. The choice of segment 150 is assumed to offset the 1D routing used in RIVMOD. RIVMOD is the IFMTFM tool to model Carson River hydrodynamics. Therefore, as RIVMOD sees the Scout Camp reach, the water table res ponse is effectively instantaneous. Comparison of Surveyed Water Surface Elev ations to Water Table Observations Surface water streamflow oscillations a nd groundwater water table oscillations are tightly coupled. With the SMS model re maining unfinished (at this point in the research) and hence unable to provide stage pr edictions, it became necessary to turn to the available water surface survey data to aid in the coupling analysis. Given the particular interest in flows at which the ba nks are contacted and susceptible to erosion, the survey data from March 2001 is the focal poi nt of this analysis, as channel flows were significantly lower during July of 2001. Water surface elevations from March 2001 ar e taken as the starting point for this analysis. Field work was conducted from March 6 to March 9, 2001. The routed streamflows for this time period are shown in Figure 6-24, with flows at Fort Churchill included for reference. While the routed streamflows for RIVMOD segments 150 through 153 are shown, it will be assumed that the fl ows through the entire model reach can be approximated by those of Segment 152 (the RIVMOD segment containing the groundwater wells). Note the unperceivable tim e lag through these segments, due to the segment length of only 250 meters; segment 150 could also have been used. During the GPS site survey, water surface elevations were collected along the water edge (in addition to the collected cha nnel bathymetry). The surveyed water surface elevations are shown in Figure 6-25. The prim arily east-west configur ation of the channel

PAGE 122

108 0.00 50.00 100.00 150.00 200.00 250.00 03/06/0103/07/0103/08/0103/09/0103/10/01 Date and TimeCarson River Streamflow (cfs) Fort Churchill segment 150 segment 151 segment 152 segment 153 Figure 6-24. Hydrodynamic routing through RIVM OD segments within reach of interest for the Scout Camp reach (Figure 5-1) allows the points to be plotted based on easting coordinate. For spatial reference, eas ting coordinates for the SMS upstream and downstream boundaries, Well 1, and extent of RIVMOD segment 152 are additionally indicated. Surveyed water surface elevations (WSEs) were classified based on the times during which they were collected, as shown in Table 6-1. Based on this classification scheme, first-cut approximations for water su rface slope were com puted with a simple linear regression. Regression equa tions in the standard form of y = mx + b, where m is the slope, with the associated R2 value, are indicated in Figure 6-25. Although these water surface slopes are not plotted based on distance along the flow path, it is evident that the water surface slope will not be constant over all flow regimes or constant over the entire model domain for a given flow regi me. Rather, water surface elevations are a function of both flow magnit ude and spatial location. The most reliable surveyed

PAGE 123

109 elevation is the indicated staff gauge. All other survey points were collected along the irregular water edge, thereby introducing human error associated with choosing the exact location of the water edge (as governed by features along the channel bed). y = -0.0013x + 2211.8021 R2 = 0.8779 y = -0.0006x + 1690.3003 R2 = 0.7664 y = -0.0013x + 2222.3785 R2 = 0.8798 y = -0.0017x + 2553.3849 R2 = 0.9244 1252 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 743800743900744000744100744200744300 Easting Coordinate (meters)WSE, NAVD 88 (meters) Q=169.6 cfs Q=172.1 cfs Q=173.8 cfs Q=212.5 cfs Staff Gage Upstream Boundary Condition Downstream Boundary Condition Well 1 Easting Coordinate RIVMOD Segment 152 Figure 6-25. Water surface elevation (WSE) as a function of flow for March 2001 data Table 6-1. Classification of surveyed wa ter surface elevations according to time of collection Survey start UBC flow at t1Survey end UBC flow at t2Average flow during time, t1(cfs) time, t2(cfs)survey (cfs) 3/6/01 15:00168.33/6/01 16:00170.8169.6 3/6/01 16:00170.83/6/01 17:00173.3172.1 3/7/01 11:00212.43/7/01 12:00212.6212.5 3/9/01 10:00173.83/9/01 13:00173.8173.8 In Figure 5-1, the flow path between cross-sections 300 and 400 is reasonably straight, contains the wells, and includes th e set of water surface elevation points for the average flow of 173.8 cfs (Figure 6-25). Looking to relate channel stage to water table elevations and being unable to justify backward extrapol ation of the surveyed water

PAGE 124

110 surface elevations from further downstream (where channel morphology is markedly different), the 173.8 cfs water surface survey set from March 9, 2001 is singled out for further analysis. This set of data points appr oximates the energy grade line for the water surface in the vicinity of the wells. With the energy grade line corresponding to energy lost along the flow path, the channel thal weg between cross-sections 300 and 400 was used to estimate the flow direction. The locat ions of surveyed water surface elevations at a streamflow of 173.8 cfs are shown in Figure 6-26, in relation to mean flow direction and well siting (axes indicate easting and nor thing coordinates). By projecting water surface observations along the direction of flow, an improved approximation of the energy grade line is obtained, as the flow path is clearly not east-west only along this portion of the channel. By plotting this set of water surface elevatio ns with respect to distance along the thalweg, shown in Figure 627, it is possible to interpolate the water surface elevation at the Well 1 cross-section (p erpendicular to flow as indicated in the figure) with use of regression analysis. The water surface nearest Well 1 is approximately 1252.30 meters (NAVD 88) for a flow of 173.8 cfs. Ideally, water table data from March 9, 2001 would be available to compare with this water surface elevation, in order to a ssess the hydraulic gradient from channel to Well 1. While the groundwater wells were installed on March 9, 2001, the pressure transducers and data loggers we re not placed in the wells un til March 12, 2001, days after the completion of the March survey work on site. In July 2001, the streamflows were between 8 and 9 cfs (lower than the levels of interest at which the banks are contacted). Stage data collected during July 2001 also cannot be compared with simultaneous water

PAGE 125

111 4503560 4503580 4503600 4503620 4503640 4503660 4503680 4503700 4503720 4503740 4503760 743920743960744000744040744080744120744160 Easting Coordinate (m)Northing Coordinate (m) Water Surface Observations Groundwater Wellsflow direction assumed parallel to channel thalweg Cross-Section 400 Cross-Section 300 Figure 6-26. March 9, 2002: location of surveyed water surface observations in RIVMOD segment 152 R2 = 0.83684 1252.20 1252.25 1252.30 1252.35 1252.40 1252.45 1252.50 020406080100120140160180200 Distance along Channel Thalweg from Cross-Section 400 (m)Water Surface Elevation (m) WSE Observations Trendline Regression Slope: 0.00098 meter per meter Location of Well 1, Perpendicular to Thalweg Figure 6-27. March 9, 2001: graphical appr oximation of channel water surface slope in RIVMOD segment 152

PAGE 126

112 table observations, even though the probes were in the wells, because the probe in Well 1 was slipping toward the bottom of the well. Th e original belief, prior to data collection, was that the water table elevat ion would basically parallel the river stage, with only a small water surface elevation change from ope n channel to Well 1 (5 feet removed from the bank). What is indeed inte resting and unnoticed in advance of this analysis, is that this is clearly not the case. Well 1 and cha nnel water levels show a tight coupling over time (based on analysis of routed streamflows), but it is here suggested that the same type of hydraulic gradient evident between Well 1 and Well 3 (at a sepa ration distance of 95 feet) also exists between the open channel a nd Well 1 over this 5-foot distance. Consider Figures 6-11 and 6-12. In these figures, the thalweg (NAVD 88 elevation = 1252.01 m) represents the lowest possible value for rive r stage, the stage which would occur at the well cross-section if the flow in the channel were essentially zero. These figures indicate that the water table observed in Well 1 is onl y situated above the thalweg elevation for flows in excess of approximately 200 cfs. At th e well transect, the thalweg is also a close approximation for the observed elevation at the base of the vertical channel bank nearest the wells. For a streamflow of 200 cfs, river stage must be above th e thalweg elevation. The SMS model is not needed to verify this point, and the need to use waders to cross the river during March 2001 provides additional assurance. The water surface elevation from open channe l to Well 1 is not a straight line, but, rather, there is a distinct hydrau lic gradient. Quantifying this gr adient was to be the use of SMS model output, but the model was not co mpleted at the time of this analysis. Simultaneous water table and stage observati ons are also not available. As the SMS model would simulate, assume that river stag e is a function of flow at any selected

PAGE 127

113 location, specifically segment 152 near the wells. Based on the tren dline from the March water surface elevations, in Figure 6-27, a ssume that the inter polated water surface elevation of 1252.30 meters (NAVD 88) is representative of a 173.8 cfs streamflow. With these assumptions in mind, this same streamfl ow magnitude occurs at other times within the data collection period for this project (March 2001 through August 2002). Antecedent streamflows, specifically rising versus falli ng stage, of course, vary. Streamflows are taken from the RIVMOD-generated hydrogr aph for segment 152. The streamflow magnitude of 173.8 cfs remained basically c onstant in the hydrogr aph during the survey hours on March 9, 2001 (Figure 6-24), the coll ection time for the water surface elevation data set under current consideration. The r outed RIVMOD hydrograph output is available at hourly intervals. Simple linear interpola tion is applied to the routed hydrograph to provide the exact times at whic h flows of 173.8 cfs occur at the well cross-section. These times can be seen in Figures 6-11 and 6-12 at the intersections of the horizontal line for Q =173.8 cfs and the hydrograph. Water table da ta is also collected at hourly intervals, but the exact times vary for each period of da ta collection (for example, 11 or 33 minutes after the hour). Linear interpol ation was performed on the well data to obtain water table observations for times corresponding to the ex act time of modeled 173.8 cfs streamflow. For times at which modeled streamflows at the wells are equal to 173.8 cfs and for which the assumed river stage is theref ore 1252.30 meters, the observed water table elevations are found to vary over an 11-cen timeter range from 1252.85 to 1252.96 meters (NAVD 88). This data set was fu rther classified based on whet her streamflow is rising or falling, with the idea that a variable time lag might exist between stage change and water table response for rising versus falling river stage. This relate s back to the possibility of

PAGE 128

114 hydraulic conductivity being impacted by considerations such as soil swelling from clay content, surface crusting after prolonged dry conditions, or entrapped air during periods of stage rising. Figure 6-28 shows points with no time lag—water table elevations at the exact time of the 173.8 cfs streamflows; note th at the x-axis indicates the times at which this streamflow occurs, and the y-axis indica tes the water table elevation. In this plot, scatter is evident, but the average water tabl e elevation in Well 1 is 1251.93 meters for both rising and falling portions of the hydrograph (dotted horizontal lines), providing some evidence of a typical stage-discharge re lationship. The estimated river stage under this flow condition is 1252.30 meters, indica ting an average head gradient from open channel to well of 37 centimeters, over a distan ce of only 5 feet. Within the data set, the minimum gradient is nearly 34 centimeters. 1251.80 1251.82 1251.84 1251.86 1251.88 1251.90 1251.92 1251.94 1251.96 1251.98 1252.00 03/01/0104/30/0106/29/0108/28/0110/27/0112/ 26/0102/24/0204/25/0206/24/0208/23/02 date and timeWater Table Elevation, NAVD 88 (meters) Rising, No Lag Falling, No Lag Average Rising Average Falling Figure 6-28. Zero time lag: temporal corre lation between river stage and water table response in Well 1

PAGE 129

115 Some of the points in Figure 6-28 are c onsidered flow “spikes,” whereby modeled flows are not really part of a distinct rising or falling trend, but only a brief oscillation. This explains why many of th e points appear close in time Removing these points, as shown in Figure 6-29, a slight increase in separation of the averag e water table between rising and falling flows is obser ved, but the separation remain s significantly less than one centimeter (note y-axis scale). Time lags were also considered. For example, streamflow magnitude is 173.8 cfs at time x, but one is in terested in the water table response 1 or 2 hours after this change. One expects the aver age difference between the mean water table observations for rising versus falling flows to increase as the time lag is increased. This is indicative of a nearly instantaneous water respon se to streamflow oscillation, or at least a response time of less than the one-hour data coll ection interval. In fact this expectation is borne out by the data. Figures 6-30 and 6-31 are for 1 and 2 hour time lags, respectively. In the figures, flow “spikes” are included, but calculations show that average rising versus falling separation dist ance increases with increasing time lag if “spike” points are removed as well. Note the elevation scale is th e same in Figures 6-28, 6-29, 6-30, and 6-31. Surface Water Modeling and Rating Curve Development Initially, surface water modeling with SM S was to predict river stage based on a realistic input hydrograph. River stage woul d be dynamically modeled. A regression analysis would then be performed to determ ine the relationship be tween river stage and water table level. By incorporating various time lags in the water table data and comparing the regression fit, two possibili ties could be assessed : 1. the indication of a positive time lag for rising flows, falling flows, or both; and 2. if present, the potential of different time lags for rising and falling flow conditions.

PAGE 130

116 1251.80 1251.82 1251.84 1251.86 1251.88 1251.90 1251.92 1251.94 1251.96 1251.98 1252.00 03/01/0104/30/0106/29/0108/28/0110/27/0112/26/0102/24/0204/25/0206/24/0208/23/02 date and timeWater Table Elevation, NAVD 88 (meters) Rising, No Lag Falling, No Lag Average Rising Average Falling Figure 6-29. Zero time lag, flow spikes removed: temporal correlation between river stage and water table response in Well 1 1251.80 1251.82 1251.84 1251.86 1251.88 1251.90 1251.92 1251.94 1251.96 1251.98 1252.00 03/01/0104/30/0106/29/0108/28/0110/27/0112/26/0102/24/0204/25/0206/24/0208/23/02 date and timeWater Table Elevation, NAVD 88 (meters) Rising, 1-hour lag Falling, 1-hour lag Average Rising Average Falling

PAGE 131

117 Figure 6-30. One hour time lag: temporal co rrelation between river stage and water table response in Well 1 1251.80 1251.82 1251.84 1251.86 1251.88 1251.90 1251.92 1251.94 1251.96 1251.98 1252.00 03/01/0104/30/0106/29/0108/28/0110/27/0112/26/0102/24/0204/25/0206/24/0208/23/02 date and timeWater Table Elevation, NAVD 88 (meters) Rising, 2-hour lag Falling, 2-hour lag Average Rising Average Falling Figure 6-31. Two hour time lag: temporal co rrelation between river stage and water table response in Well 1 Justification for a change in strategy Ultimately, the modeling limitations of the SMS model required an adjustment to be made to this original approach. Variable -sloped banks, a steep inner-channel thalweg, and irregular bed and bank mo rphology features resulted in jagged wet / dry boundaries and recurrent loss of model stability b ecause of divergence problems. Divergence problems resulted in extended model run time model crashes, and invalid solutions beyond reasonable ranges for velocity or dept h. An iterative, time-consuming procedure involved with mesh refinement and editing, accompanied by the strategic manipulation of model parameters, required se veral months of effort to overcome these obstacles and enable the model to simulate a fairly simp le lowering of the wate r surface elevation.

PAGE 132

118 Eventually, in March 2003, the SMS mesh was deemed complete and was capable of simulating a lowering water surface from bank-full conditions to almost zero-flow conditions, subject to a head gradient of 0.2 meters from the upstream to downstream boundaries of the mesh. It was necessary to abandon the simulation of overbank stage levels and real-time dynamic streamflow hydrographs within th e SMS model domain. Experience with the FLO2DH model within SMS indicated that el ement re-wetting could not be simulated along the Scout Camp reach without the m odel experiencing a crash or divergence problem causing an invalid solution. Only th e falling portions of the routed hydrograph could be simulated with accurate regard for antecedent flow conditions. Recalling that no rating curve relationship is available at the downstream boundary, no means are available to use the SMS mesh subject to an upstr eam flow boundary condition. Use of a flow upstream boundary would require specification of a downstream rating curve; of course, even if rating curve data was available, th e rating curve option w ithin SMS was found to not function properly. Instead, head boundary conditions were used to gradually lower the water surface elevations across the model domain, with flow values computed from depth and velocity output along nodestrings Mesh editing was completed using a conservative head gradient of 0.2 meters, but this va lue was not based on any known data. The limited availability of water surface elevation survey data prevented model calibration or verification, and the flaws of using a constant head gradient between the model flow boundaries should be noted. Most notably, stage-discharg e relationships are assumed independent of antecedent flow c onditions, and head loss from upstream

PAGE 133

119 boundary to downstream boundary is assumed cons tant over the entire range of modeled flow magnitudes. The presence of a low-flow inner channel, a channel bed characterized by erosion and deposition features, channel bank s of variable slopes ranging from flat to vertical, and rapid zones of sl ope transition in both longitudi nal and lateral directions along the channel virtually guarant ees that these will not be the case in reality. That said, the assumptions remain necessary given the intrinsic data limitations. By the time the SMS model was fully operational, alternative approaches to a regression analysis of modeled river stage versus water table elevation had been used to circumvent the use of the SMS model as a component for regression analysis. By comparing routed streamflow hydrographs to continuous water tabl e observations and making use of the limited water surface survey data that is available, it was shown that river stage and water table le vels are tightly coupled, and the water table response to streamflow oscillations is effectively inst antaneous. No evidence of a time lag on the order of hours was indicated for either rising or falling flow conditions. This conclusion was used (chapter 7) for the design of bank elements and determination of bank element functionality. Note that the same questions originally to have been answered through regression analysis (and making use of SMS) were answered with these alternative analysis approaches. No utility is found for a regression analysis of SMS stage predictions versus water table observations. To establ ish utility to the months spent developing the SMS model, the decision was made to use the model outpu t to generate a rating curve at the well cross-section and compare rating curve stage predictions to surveyed water surface data. The 0.2-meter head gradient used for mesh re finement was arbitrary and selected to be

PAGE 134

120 conservative, but model operation does require user specification of a head gradient. Determination of an applicable gradient to be used in rating curve development for the Scout Camp reach is the subj ect of the next section. Determination of FESWMS longitudinal hydraulic gradient The decision was made to simulate lowering of the water surface and element drying subject to a physically-m eaningful head gradient. The model parameterization, as reviewed in Appendix A remained unchanged. Simulations were to range from bank-full conditions to the minimum level the mesh c ould handle. Uniform flow is considered. Under conditions of uniform flow, one assu mes the water surface slope, energy grade line slope, and bed slope as equal. Based on survey data, and, as would be expected given the bend internal to the modeled reach, the bed slope is not uniform between any consecutive cross-sections. Only two boundary conditions ar e input into the SMS model, eliminating the possibility of user specification of multiple water surface slopes within the model domain. Instead, a representative bed sl ope for the 500-meter modeled reach was determined using regression analysis. This sl ope is based on surveyed thalweg elevations along the channel, as extracted from the co mplete set of surveyed data points. The thalweg is a connection of the poin ts of lowest elevation for all crosssections, representing minimum flow level in the channel. The latitude and longitude coordinates of all points along the thalweg are shown in Figur e 6-32, as is their location relative to other points within the surveyed reach. Linear segments were used to connect the data points and compute a distance along the thalweg, or dist ance along the flow direction, for each point. The thalweg point located along cross-section 500 (Figure 5-1) is used as the point of zero distance. By performing a regre ssion of bed elevation versus

PAGE 135

121 4503500 4503550 4503600 4503650 4503700 4503750 4503800 4503850 743800743900744000744100744200744300744400 EastingNorthing Survey Points: March, July 2001 Thalweg Figure 6-32. Channel thalweg location base d on survey points from March and July, 2001 R2 = 0.895 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 0.0100.0200.0300.0400.0500.0600.0 Distance Along Thalweg from Upstream Boundary Cross-Section (meters)Surveyed Elevation, NAVD 88 (meters) July Thalweg Trendline Downstream Boundary of SMS Model Domain Upstream Boundary of SMS Model Domain Regression Slope: 0.00391 meter per meter A ssume Uniform Flow (SEGL = Sws = Sb) Surface Water Head Loss over SMS Model Domain = 1.99 meters 0.00391 Figure 6-33. Linear regression of March, July 2001 surveyed bed elevations for use in determination of hydraulic gradient for SMS model

PAGE 136

122 distance along the thalweg, the channel is e ffectively straightened, and a mean bed slope is computed. The results of the regression analysis can be found in Figure 6-33. The figure shows that bed slope is variab le over the model domain, but the R2 value of 0.895 does indicate a good fit and a definite tre nd for elevation loss moving downstream. Regression analysis, based on the slope of the trendline in Figure 6-33 (S = 0.00391 m/m), indicates a net elevation loss of 1.99 me ters through the 510-meter flow path of the Scout Camp model domain, from crosssection 500 upstream to cross-section 100 downstream. Cross-sections 100 and 500 coinci de with the upstream and downstream boundary nodestrings with in the SMS mesh. Hydrodynamic simulation and rating curve development using FESWMS Based on the assumption of uniform flow the head gradient across the model domain was assumed equal to the mean bed sl ope. The 1.99-meter head gradient does not mandate a uniform water surface slope within the mesh interior, as the SMS model will perform these computations based on the mesh geometry. Visual inspection of the points’ locations relative to the regression line indi cates that the SMS model will predict smaller flow depths in the central portion of th e modeled reach and larger flow depths downstream. This observation is consistent with conditions observed in the field, and also suggests eventual flow discon tinuity within the SMS mesh interior, indicating a minimum attainable level for flow simulations. It is no ted that this is not a perfect solution because water surface slope is not cons tant with respect to flow or location, but rather the assumption is a product of limited data. This wa s previously suggested from the analysis of surveyed water surface elevations in Figure 6-25. Complex, dynamic channel geometry and lack of velocity observati ons prevent artificial development of a downstream rating curve using a relationship su ch as Manning’s equation. Additionally,

PAGE 137

123 the single known accurate water surface elevatio n data point from the staff gauge, is not sufficient for use in SMS calibration for the channel bed and channel bank roughness coefficients. Rather, the bed and bank roughness values are assumed equivalent, with this value equal to the 0.045 value used in RIVMOD. Notably, this value is likely set high because of the 1D modeling approach embodied in RIVMOD. The head gradient across the model domain was used to develop a rating curve at the well cross-section. The SMS model was initialized for the 1.99-meter head gradient with all nodes initially subm erged. After the initial wate r surface was lowered to 1255.0 meters at the upstream model boundary a nd 1253.01 meters at the downstream boundary (note head gradient equal to 1.99 meters), ra ting curve computations were initialized. The water surface elevations were simultaneous ly lowered at both boundary nodestrings, at the same rate, to maintain a constant head gr adient. To maintain model stability and limit output file size, the boundary conditions were lowered continuously by a series of simulations and hotstarts. When the water su rface could no longer be lowered because of flow discontinuity within the mesh interior (wet elements are no longer connected), the simulation was necessarily stopped. Examples from the time series of element drying due to the lowering water surface are shown in Appendix C including the minimum attainable boundary condition levels. After the water surface was lowered to minimum attainable levels, the intermediate solution files were used as starting points for a series of steady state simulations. Hotstart files were used to st art these steady state c onditions at 0.25-meter head increments (chosen to generate a suffi cient number of data points). During mesh construction, nodestrings were constructed along cross-sections of the mesh interior, to

PAGE 138

124 be used for obtain streamflow as a func tion of head boundary conditions. Steady state simulations were run consecutive ly using hotstart files until the following criteria is met: flow across the upstream boundary and well cr oss-section nodestring s have stabilized (flows may not be exactly the same as flow direction is not exactly perpendicular to the well nodestring, but they will be very close; range was later determined as 0.15 – 0.60 cubic meters per second (cms)); and the water surface elevation at th e well cross-section node along the vertical bank adjacent to the wells is no longer changing between simulations. With jagged wet / dry boundaries within the model domain, convergence criteria were difficult to attain. The working definition for attainment of steady state was therefore necessarily adopted. At the well cr oss-section, steady state flows and water surface elevations were recorded after each simulation for later use in rating curve development. Boundary condition increments of 0.25 meters were used until minimum flow levels were reached. The steady state outputs from the SMS model are summarized in Table 6-2 (UBC / DBC = upstream / downstream boundary condition). Table 6-2. Steady-state output for depth and discha rge from final SMS model Equilibrium Flow at Well TransectWater Surface Elevation Adjacent to Well 1 (GZF = 1252.01 m) UBC Stage, DBC Stage, NAVD 88 (m)NAVD 88 (m) Q1 (cms)log10 Q1Q (cfs)NAVD 88 (meters)Depth (above GZF, m) log10 (Depth) 1255.001253.01226.802.3680141254.502.490.40 1254.751252.76190.202.2867211254.282.270.36 1254.501252.51152.202.1853781254.052.040.31 1254.251252.26109.602.0438731253.961.950.29 1254.001252.0184.201.9329751253.741.730.24 1253.751251.7663.701.8022511253.491.480.17 1253.501251.5145.301.6616011253.231.220.09 1253.251251.2620.801.327351253.141.130.05 1253.001251.0115.201.185371252.810.80-0.10 1252.751250.766.850.842421252.450.44-0.36 1252.5251250.540.80-0.10281252.400.39-0.41 The upstream and downstream stage boundary conditions were input to the SMS model based on the 1.99-meter mean gradient determined from regression. With the SMS model parameterized in metric units, the streamflow / di scharge (Q) through the well cross-section and water surface elevation for the node along this transect closest to Well 1

PAGE 139

125 are extracted from steady-state model out put in units of cms and meters (NAVD 88 vertical datum), respectively. The procedure used for rating curve development was based on techniques practiced by the USGS (Kennedy, 1984). The stag e that would occur at a gaging station if the discharge were infinitesimal is the gage height of zero flow (GZF). For this analysis, the GZF was defined as the elevation of the channel thalweg (the minimum bed elevation) along the well transect, a va lue of 1252.01 meters (NAVD 88 datum). The stage versus discharge relations hip is shown in Figure 6-34, w ith each point taken from a steady-state SMS model si mulation in Table 6-2. 1251.5 1252.0 1252.5 1253.0 1253.5 1254.0 1254.5 1255.0 1255.5 0255075100125150175200225250 Discharge at Well Transect (cms)WSE Predicted by SMS, NAVD 88 (meters) Figure 6-34. Stage – Discharge Relationshi p for Steady State Lowering of SMS Water Surface Elevation Rating curve analysis is base d on fitting a straight line through the stage-discharge relationship in log-log space. Stage is first normalized to depth (above the GZF datum). The log transforms of flow and depth data points are incl uded in Table 6-2. Before

PAGE 140

126 determination of a rating curve equation, limitations of the data were noted. First, the top of the channel bank is situated at an elevat ion of 1254.15 meters. Stage elevations greater than this value correspond to flows on the floodplain, beyond the specified upper limit (for accurate stage modeling) of the constructed SMS mesh. It is also evident from Figure 6-34 that the data point corresponding to the th ird-highest discharge (152.2 cms) does not follow the trend observed for lower discharge values. Given the irre gular nature of topof-bank elevations along the cha nnel, this data point was al so presumed to correspond to floodplain flows for some areas of the m odel domain. Recalling that the maximum observed flow between March 2001 and Septem ber 2002 is less than 1500 cfs (42.5 cms), the range of flow magnitudes for the period of interest will be entirely included by regression of the remaining points. Additionally noted is the fact that the final data point in Table 6-2 does not correspond to the 0. 25-meter boundary condition lowering interval. The data point noted at a flow of 0.80 cm s, as shown in Figure 6-34, is the lowest available modeled stage for which flow is able to be transported through the model domain. At this flow level, the water surface elevation immediately adjacent to the wells is disconnected from the actively flowi ng portion of the channel, representing a backwater location, and this data point was therefore also excluded from the rating curve analysis. The lowest streamflow for whic h elements adjacent to this bank remain hydraulically connected to the active flow path is 1.91 cms. The applicable flow range for the rating curve equation was therefore define d as 1.91 cms (67.5 cfs) to 109.6 cms (3871 cfs). The rating curve analysis included the 7 middle data points highlighted in Figure 634. The log transforms of these data points we re plotted and fitted with a regression line. Regression results are found in Figure 6-35. The regression line is extrapolated back to

PAGE 141

127 show the complete applicable discharge rang e. The stage-discharge relationship predicted from this regression line is s hown in Figure 6-35. The techniqu es to generate this curve are next discussed. The three largest flow magnitudes represent floodplain flows outside the rating curve applicable flow range, so th e poor match to these data points is expected. y = 1.9182x + 1.432 R2 = 0.9413 0.00 0.50 1.00 1.50 2.00 2.50 -0.60-0.50-0.40-0.30-0.20-0.100.000.100.200.300.40 log10 (depth above GZF in meters)log10 (discharge in cms) SMS Output Trendline Figure 6-35. Linear regression for determin ation of discharge as a function of depth The straight-line fit th rough the data points assu mes a simple power law relationship for discharge as a functi on of depth above the GZF datum: D Q (Eq. 6-8) where Q is discharge (cms), D is normalized depth (meters), and and are constants computed using the straight-line regression in Figure 6-35. The constants are determined directly from the slope and intercept of th e regression equation, as determined from equation 6-9: log10 (Q) = log10 + log10 D (Eq. 6-9)

PAGE 142

128 This equation has the form of a straight lin e: y = mx + b. Plotting the log of normalized depth on the x-axis and the log of flow on the yaxis, it is noted that the y-intercept of the regression line corresponds to log and the slope of the regr ession line corresponds to The values of the constants are computed to be the following: 040 27 10432 1 918 1 The power law rating curve function for the Sc out Camp well cross-section is therefore described by equation 6-10: 918 1040 27 D Q (Q in cms, D in meters above GZF = 1252.01) (Eq. 6-10) An inverse relationship for depth as a functi on of flow (flow being the known parameter from RIVMOD routing through the well reach) can be calculated in several ways. A power law relationship is computed in the same manner as above by switching the axes on which discharge and depth are plotted, as in Figure 6-36. Without showing the details, the power law relationship for depth (above GZF) as a function of discharge is described by equation 6-11: 491 0200 0Q D (Q in cms, D in meters above GZF = 1252.01) (Eq. 6-11) The power law relationship, as previously discussed, in only applicable for the streamflow range between 1.91 cms (67.5 cf s) to 109.6 cms (3871 cfs). The groundwater well transect is approximatel y centrally located within IFMTFM segment 152. Making use of the streamflow hydrograph through th is segment as routed by RIVMOD and the relationship for depth as a func tion of streamflow (equation 611), plots were prepared to show SMS-modeled river stage versus water table levels in Well 1, noting the 5-foot separation distance between channel and well. For routed streamflows within the

PAGE 143

129 y = 0.4907x 0.6994 R2 = 0.9413 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 0.00 0.10 0.20 0.30 0.40 0.000.501.001.502.002.50 log10 (discharge in cms)log10 (depth above GZF in meters) SMS Output Trendline Figure 6-36. Linear regression for determin ation of depth as a function of discharge acceptable magnitude range, water surface eleva tion along the well transect is calculated for comparison with observed water table r ecordings from Well 1. As indicated in Figures 6-37 and 6-38, a positive hydraulic gradient ranging between 0.4 and 0.6 meters exists between the water surface elevation in the channel (adjacent to the channel bank) and the water table observation in Well 1. It is important to compare the rating curve predictions with the limited field data available for water surface elevation in the chan nel, in the vicinity of the wells. In Figure 6-27, a regression equation approximated the wa ter surface elevation at the well transect as 1252.30 meters (NAVD 88). The water surf ace elevation survey data used was collected on March 9, 2001 for streamflow conditions of 173. 8 cfs. Using equation 6-11 or by inspection of Figure 6-37 on this date, it is noted that the rating curve predicts a water surface elevation of 1252.45 meters at this flow magnitude. The water surface

PAGE 144

130 estimates from these two approaches ar e not in good agreement, differing by 15 centimeters. This result was not unexpected. Gr eater reliability is placed on the estimate based on the surveyed elevations for four ma in reasons: the surface water model used to develop the rating curve is not properly calibrated based on the lack of available data; the rating curve calculations assume streamflows routed (from FCH to the Scout Camp well transect) using RIVMOD to be correct (these are the discharge values plugged into the rating curve equation), although RIVMOD does not consider losses to ET or groundwater; the bed roughness used in the quasi-3D SMS model is set equal to the value used in the 1D RIVMOD, possibly causing mode l stage predictions to be higher than for a lower bed roughness value; and the SMS mode l assumes a head grad ient that does not vary over time, equal to the 1.99-meter averag e bed slope. With these considerations in mind, the 15-centimeter difference was reason ably expected. Unfortunately, additional comparisons for alternate flow regimes are not possible without additional water surface elevation data. However, it is important to no te that the water surface elevations predicted from SMS output with this rating curve a pproach (Figures 6-37 and 6-38) in fact represent maximum water surface elevations that could be observed in the channel, based on the above-listed assumpti ons. Alternative model runs for different bed roughness values are not completed given poor data av ailability regarding the setting of this parameter, the required model run time of se veral days, and the inc onsequential value of these results on the conclusions reac hed for bank element functionality. Water surface oscillations app ear to precede water table os cillations in figures 6-37 and 6-38. With previous evidence presented of a hydraulic gradient away from the

PAGE 145

131 1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5 3/7/20014/6/20015/6/20016/5/20017/5/20018/4/20 019/3/200110/3/200111/2/200112/2/20011/1/2002 date and timeElevation, NAVD 88 (m) Water Surface Elevation Well 1 Water Table Elevation Figure 6-37. 2001 rating curve-based estimates for water surface elevation in the channel compared to water table level in Well 1 1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5 1/1/021/31/023/2/024/1/025/1/025/31/026/30/027/30/028/29/029/28/0210/28/02 date and timeElevation, NAVD 88 (meters) Water Surface Elevation Well 1 Water Table Elevation Thalweg at Well Cross-Section Figure 6-38. 2002 rating curve-based estimates for water surface elevation in the channel compared to water table level in Well 1

PAGE 146

132 channel, this response should be reversed. Recall, this same problem was encountered when streamflow oscillations for segments 151 or 152 were compared with water table oscillations. This is the result of the 1D ch annel routing approach used by RIVMOD. The common thread between the tw o comparisons is that the same streamflow hydrograph from RIVMOD segment 152 was used with th e rating curve equation to predict the plotted water surface elevations shown in Figures 6-37 and 6-38. Completion of a more detailed regressi on analysis of water surface predictions from SMS versus water table oscillations s een in the well was not warranted and would prove meaningless without accurate routed hydrographs and additional data for use in model calibration for the Scout Camp reach. Ba nk storage, evaporation, and groundwater losses downstream of Fort Churchill are not considered within RIVMOD. The surface water model is not properly calibrated based on the absence of sufficient water surface data, and, as such, the rating curve com putations are only ap proximate. Sufficient evidence has been presented to justify the assumption of an instantaneous saturation height response to streamflow changes with in the erodable face of the channel bank, and this evidence was the basis for bank element design. Limitations of FESWMS and impa cts on this thesis research The SMS modeling package was purchased fo r this thesis research based on its reasonable cost and stated ability to simula te areal wetting and drying for a riverine system, including the capability to handl e steep channel banks. As thoroughly documented, model development for the Sc out Camp reach of the Carson River was neither straightforward nor simple. Use of th e SMS software began in June of 2002, and a useable model was not completed until March 2003. The SMS model package includes two possible modules for 3D hydrodynamic modeling: RMA 2 and FESWMS. FESWMS

PAGE 147

133 was ultimately chosen based on its support of both subcritical and supercritical flow analysis. The high degree of variability in el ement slopes and river stage levels create a high likelihood of supercritical flow cond itions at the mesh boundary conditions and within the mesh interior. In fact, this premonition was s upported by supercritical flow warnings encountered during many model simulations. The FESWMS module and the FLO2DH code possess some notable limitations. Recall, when this thesis research began, one was looking to use FESWMS to simulate a continuous hydrograph through the surveyed re ach and extract water surface elevation data. The model shortcomings are briefl y reviewed. First, a downstream boundary condition must be specified, rather than computed by the model. With no available downstream rating curve, the model could onl y be run under conditions of a constant longitudinal head gradient for all flow regime s, despite field data indicating this as an invalid assumption. This is also a result of a distant field site (Nevada) and a limited ability to collect detailed water surface data over time. The mesh ge neration process does not favor the use of detailed survey data over a relatively small reach. In this instance, the high node density, with typical spacing of 3 to 5 meters, and irregular nature of the channel morphology causes extreme divergence w ith the finite difference algorithm used by FLO2DH. The surface variability, due to er osion and deposition cycl es, is present in the field and therefore enters into the mesh Large irregularities understandably cause divergence problems, but localiz ed variability on the orde r of a couple of centimeters should be better handled within the model. Ideally for the model, data points (model nodes) would fall along contour lines for im proved drying performance; and spatial heterogeneity, such as highly variable ba nk slopes ranging from gradual to vertical,

PAGE 148

134 would not exist. The fact remains, however, th at in order to devel op an accurate water surface solution within the SMS model, high sp atial resolution is needed (as opposed to perhaps a simple linear interpolat ion between cross sections). Steep channel banks cannot be simulate d by the FESWMS mode l, unless vertical wall boundaries are used. This el iminates the model ability to simulate floodplain level flows. The transition zones from steep vertical banks to gradual sloping banks within the model domain also required a large degree of mesh editing. Divergence was also a problem when multiple material types were us ed to indicate variable roughness values for bed, banks, and floodplain areas. While the to ols for mesh editing are acceptable, mesh quality indicators, warning messages, and e rror explanations coul d be improved upon to better identify potential problem areas. Typical of many models, the available documentation, although extensive, is va gue, and the physical meanings of model parameters are often unexplained. Technica l support through ems-i, the company through which the SMS model was purchased, should also not be commended. The most substantial model shortcomings relate to wetting and drying. The model would be suitable to simulating modest stage fluctuations across a la rge river area, but the model is not very useful for simulating la rge stage oscillations and repeated element wetting and drying across a small river area (i .e. the Carson River). The parameters to determine if an element is wet or dry— storativity and depth tolerance for drying— perform poorly when some mesh elements are steep and element slopes are highly variable, even across short distances. The end result for the Scout Camp mesh is a jagged wet / dry boundary. When the boundary ta kes this jagged shape, convergence requirements are seldom attained and model run time is increased. The jagged boundary

PAGE 149

135 also increases the lik elihood for element re-wetting, w ith the usual end result a model crash. With no guidance available for setting these parameters in model documentation, a discernable fine line exists between experime nting with these parameters to accurately predict the wet / dry boundary (likely causing a model crash) and loosely setting these parameters to eliminate element re-wetting (thereby bringing into question the accuracy of the wet / dry boundary and the solution wate r surface elevation). Priority was given to accurately simulating the wet / dry boundary wh en possible, with the trade-off being no ability to simulate a rising river stage. Lastly, a comment on model simulation time is needed. Model simulations must begin with all mesh nodes submerged. This is a waste of time during the mesh editing process. When a node is identified as causi ng a model failure or divergence, the mesh must be edited. Even if only one node is alte red, all previous solution data for the other 12000+ nodes must be discarded and the model simulation started again. By attempting to minimize the magnitude of elevation corre ctions (to preserve the original field measurements), it is never certain that a given correction will be successful until the water surface reaches the previous problem level. Based on limited explanations regarding the cause of a model crash or divergence, mesh editing is iterative with only a limited number of problem nodes able to be corrected following a given simulation. When starting with all nodes submerged, th e rate at which the water surface can be lowered serves as a further constraint, as m odel stability must be maintained throughout the process. By editing the so lution file, nodes can be manually turned off (forced to be dry). This is particularly us eful to manage disjoint wet elements caused by jagged wet / dry boundaries and element re-wetting. After th e solution file is edited, the simulation

PAGE 150

136 can be hotstarted from this edited solution, and a model crash can be averted (thereby saving the need to start over completely). The SMS model would be better suited to modeling large riverine systems from coarse bathymetry data. The spatial heteroge neity of Carson River bathymetry, vertical banks, and bed features proved to be signi ficant obstacles for the software program. Conceivably, based on this user’s experien ce with the SMS model, particularly the FESWMS module, the program would work well to simulate a river system with greater bathymetric uniformity and could simulate sma ll-scale wetting and drying in which only a small portion of the model mesh is under going wetting and drying. The channel size and streamflow variability of the Carson Ri ver was a poor match for the SMS software package.

PAGE 151

137 CHAPTER 7 BANK ELEMENT DEVELOPMENT AND COMPARISON TO PREVIOUS IFMTFM General Considerations for Bank Element Design and Operations Methylation within channel banks is depe ndent on the presence of anoxic regions within the sediments. The size of the satura ted, anoxic zone of channel bank sediments changes with river stage, in response to th e induced hydraulic gradient away from the channel. Bank sediments which are characte rized by these conditions long enough for the microbial populations to become establishe d are able to support an active zone of methylation and demethylation. By tracking the size of the ac tive zone of methylation as river flows change over time, the updated IFMT FM is able to remove itself from the unrealistic assumptions of static MeHg con centrations within the channel banks and no MeHg above a given depth threshold. By allowing the model to run for a sufficient length of time prior to generating used output, in itial conditions (bank concentrations) can effectively wash out of the system, there by removing the dependence of output on initial conditions. The spatial heterogeneity of Carson River sediments proved to be an obstacle for detailed unsaturated zone and groundwater mo deling. Visual inspection of soils on site verified sediment heterogeneity, as did the pump test results indi cating the presence of high conductivity layers. To develop and valid ate a 2D groundwater and unsaturated zone model, detailed soil data is required. Th e modeling of flow in the unsaturated zone requires user specification as to how moisture content, relative conduc tivity, and specific moisture capacity vary as a function of pr essure head. Ideally, these curves are

PAGE 152

138 determined directly by performing a series of tests on the soils involved in the study. In many cases, they can be approximated using a set of measured or approximated constants and a set of empirical relationships, an exam ple being the van Genuchten functions (Lin et al., 1996). In fact, one w ould ideally possess data relati ng channel bank capillary fringe characteristics to the rate at which an adj acent head boundary (river stage) is rising or falling. The combined vertical extent of the partially saturated zone will depend on whether river stage is rising or falling. A sma ller distance would be expected if the river stage is rising (bank not recently inundated) a nd a larger distance if river stage is falling (bank recently inundated). In the end, time constraints and the conclusion that such detailed sediment data could not justifia bly be extrapolated over the IFMTFM model domain prevented this type of analysis from being undertaken. Analysis of the coupling of routed stream flow and water table response; surveyed water surface elevations; and pump test data indicated a nearly instantaneous response time between river stage fluc tuations and observed water table changes in Well 1. A hydraulic gradient away from the river channel is evident for periods of river contact with channel banks within the modeled reach. Given the nearly instantaneous response observed at a distance of 5 feet into the ch annel bank (location of Well 1), it is assumed that the saturated portion of the erodable face of the channel bank is in instantaneous equilibrium with the modeled river stage, as predicted within the IFMTFM by RIVMOD. In order to account for the microbial respons e time following a change in bank moisture, the saturated portion of the bank is tracked over time. The IFMTFM will consider the portion of the channel bank ac tively methylating to be a temporal average of this saturation height. The determination of this averaging window is later discussed.

PAGE 153

139 Conceptually, however, if the river stage is rising over a few days time (time must exceed the averaging window), the actively methylati ng portion of the bank must be lower than the current river stage. Similarly, if river stage is falling for a consecutive number of days, the actively methylating portion of the ba nk will be higher than the final river stage. Erosion only occurs for portions of the bank contacted by the river, such that MeHg generated within the channel bank above the river stage will only be introduced to the water column if the stage again rises to contact these bank areas prior to the bank becoming dry. Within the model, microbial methylation and demethylation cease when bank sediments become dry. If the sediments are re-wetted, the sediments again require the same acclimation time for microbes to become established, regardless of how recently the sediments were saturated. The use of a temporal averaging window to track the saturate d portion of the bank requires the creation of bank elements within the IFMTFM. In following the original goals of this thesis, the ability to be tter simulate field processes by allowing methylmercury bank concentrations to change over time was added to the previous capabilities of the IFMTF M. Model calibration requirements are also reduced, as discussed in the upcoming sect ions of this chapter. Updated Input Data Sets to Include Bank Elements Model Functionality Prior to Addition of Bank Elements The complete IFMTFM model includes 274 st ations, with station 1 located at the Carson City Gage (CCG) and station 274 at Lahontan Dam. Channel cross-sectional geometry is specified in the RIVMOD input data file, based on the parameters defined in Figure 3-1. Linear interpolati on of channel geometry is a pplied to segments, or river reaches between adjacent stations. Prior to modi fications as part of this thesis research,

PAGE 154

140 the IFMTFM included two element types, water column elements and bed elements. Segments 1 through 273 are classified as water column elements, and segments 274 through 546 are classified as channel bed elements. Each water column segment lies above a corresponding bed segment, as indicated in Figure 7-1. Bed segments encompass the channel bottom area, intern al to the main channel (BW2 in Figure 3-1) between adjacent stations. Mercury methylation and de methylation within channel bed elements create dynamic mercury concentrations able to exchange with the overlying water column through sediment erosion and diffusi on processes. The active depth for mercury transformations within channel bed elements is assumed to be 10 centimeters. Segments 1, 274 Segments 2, 275 Segments 3, 276 Segments 4, 278 WATER COLUMN SEGMENTS BED SEGMENTS STATION 1 STATION 2 STATION 3 STATION 4 STATION 5 Figure 7-1. Station and segm ent definition within IFMTFM Within the previous IFMTFM, channel banks (vertical portion of D2 in Figure 3-1 that is above the slope SL1) can be eroded (although geometry is not updated within the model), but mercury transformations do not occur within the channel banks. Rather, inorganic Hg(II) and organic MeHg concen trations within channel bank sediments remain constant, as determined by calibration on the parameters 1 and 2. 5 0 0 2S Hgbank MeHg (Eq. 4-1)

PAGE 155

141 5 0 0 1] [ S Hgbank in (Eq. 4-2) To model dynamic mercury concentrations with in the channel banks, a new segment type was needed. Bank segments, incorporating th e channel bank between adjacent stations, were created within the IFMTFM hydrodynami c input data file. These new segments within IFMTFM will be collectivel y referred to as bank elements. The hydrodynamic portion of the IFMTFM (RIVMOD) predicts water depths and velocities for all segments throughout the Carson River model domain. Based on the specified cross-section geometry for any tw o adjacent stations, the model is able to compute the average contacted height of th e water column on the channel banks for all segments. By tracking this height over time, the model estimates the saturated portion of the channel banks in each of the 273 water co lumn segments. Within the saturated bank sediments, methylation and demethylation are modeled to occur. MERC4 modeling of mercury partitioning to bank sediments and por e water is used to incorporate diffusion and advection of mercury from the channel ba nks to the water column, in addition to the already-modeled bank erosion mechanism (unchanged). Bank element functionality is essentially the same as bed element functiona lity with respect to mercury modeling. Rather than starting with the complete 273-segment IFMTFM model of the Carson River, a simple 10-segment model was constructed. Modifications to the IFMTFM code and input data files were fi rst tested on a small scale with a smaller computation time. The simple model included 10 water column segments, 10 bed segments, and 10 bank segments; uniform cross-section geometry; and a uniform bed slope. Conceptually, the updated IFMTFM model was constructed as shown in Figure 7-2. Arrows indicate mercury and sediment exchange between the different element types

PAGE 156

142 within the model for any given river segm ent, based on average values between two adjacent stations. Due to the geometric symmet ry used to define all model segments, one bank segment corresponds to a single water column and bed segment for each model reach between adjacent stations. Total bank segment volume over a reach is simply the predicted bank volume for the single bank segment multiplied by two (left and right banks). This same approach is used to compute surface area for diffusion. Figure 7-2. Conceptual IF MTFM segment interaction Modified Input Data Files Following the code modifications to the IF MTFM and testing and refinement based on the simple 10-segment model, updated input data files for the hydrodynamic (RIVMOD) and water quality (WASP5/MERC4) components of the IFMTFM were prepared to match the new requirements of the code. Based on the linkage structure between the three IFMTFM models and the unc hanged channel geometry specified in the RIVMOD input data file modifications are only necessary for the WASP input data file. The required changes are briefly discussed in their order of appearance within the WASP5 model documentation. Data Group A: Model Identific ation and Simulation Control o 819 model segments: 273 water column segments, 273 channel bed segments, and 273 channel bank segments Water Column Segment Channel Bank Segment Channel Bed Segment

PAGE 157

143 Data Group B: Exchange Coefficients o 818 modeled exchanges: 272 water column segment to water column segment, 273 water column segment to bed segment, and 273 water column segment to bank segment o Characteristic length for diffusion for ba nk segments set equal to 0.01 m value used for bed and bank segments o Diffusion coefficient for bed and bank segments changed to 1.14E-10 to match field data for mercury flux out of channel bed Data Group C: Volumes o Defined new segment type, ITYPE = 5 for bank segments o 273 bank segments numbered as segments 547 through 819 o Bank segments defined based on adjacent water column segments (segment 1 with segment 547, segment 2 with segment 548, etc.) Data Group G: Parameters o Three additional segment classifica tions added, including river bank segments, upstream reservoir bank segm ents, and downstream reservoir bank segments o Specified bank segment parameters same as corresponding bed segment parameters, with exception of methylation and demet hylation rates (further discussed under Data Group I) Data Group I: Kinetic Time Functions o Model limitations dictate a maximum of 4 methylation / demethylation rates based on segment classification Previously, rates varied for water column segments, river bed segments, upstream reservoir bed se gments, and downstream reservoir bed segment After changes, rates vary for water column segments, river bed segments, reservoir bed segments (average of previous upstream and downstream reservoir values), and bank segments Data Group J: Initial Concentrations o Initial mercury and sediment concentr ations specified for bank segments Hg(0) concentrations = 0.00 (not simulated) Hg(2+) concentrations = 1.00E+01 (a rbitrary as later calibrated on) MeHg concentrations = 0.00 (no ava ilable data, methylation simulated by updated IFMTFM model and concentrations updated) Hg(inert) concentrations = 0.00 (not simulated) Coarse suspended solid (Solid 1) concentrations based on sediment samples from river banks and rese rvoir bed (total conc.=1.56E+06) River segments 547-755: CSS = 1.08E+06 (69.5% of total)

PAGE 158

144 Reservoir segments 756-819: CSS = 0.62E+06 (40% of total, same as corresponding bed segments) Bedload (Solid 2) concentrations = 0.00 (bedload only present for water column segments) Washload (Solid 3) concentrations based on sediment samples from river banks and reservoir be d (total conc.=1.56E+06) River segments 547-755: WL = 0.48E+06 (30.5% of total) Reservoir segments 756-819: WL = 0.94E+06 (60% of total, same as corresponding bed segments) No changes necessary for Data Groups D: Flows, E: Boundary Concentrations, F: Loads, and H: Constants Comparison of Updated IFMTFM Code versus Previous Version Code modifications were completed by J ohn Warwick in 2003. While inspection of the final code changes are not needed for th is discussion, it is important to understand how the updated code incorporates the new bank elements for improved mercury modeling; and the resulting be nefits to model calibration. Throughout this discussion, the updated IFMTFM code with bank elements (‘n ew’ version) will be compared to the previous version of the model code without bank elements (‘ old’ version). Changes to the IFMTFM by this thesis research only pertai n to mercury modeling within the channel banks. Remaining model structur e (chapter 4) was unchanged. Mercury Modeling within ‘Old’ IFMTFM Within the IFMTFM, the MERC4 model simulates inorganic Hg(II), methylmercury (MeHg), and the following th ree solids types: washload, CSS (coarse suspended solids), and bedload. Note that only the soluble forms of these Hg(II) and MeHg are assumed available for chemical transformation. The simulation of solids is critical because of the strong tendency of mercury to adsorb onto solids. The MERC4 model resulted in the coupling of chemical speciation and kinetics with the transport capabilities of the WASP4 modeling system. A brief overview of sediment transport

PAGE 159

145 modeling within the IFMTFM was discussed in Chapter 3 of this document. Documentation of erosion and floodplain depo sition modeling are available from Carroll et al. (in press). The details of these modeling processes are not here reviewed as they remain unchanged throughout the course of this research. Additional details concerning mercury modeli ng within the IFMTFM are critical to the understanding of bank element functiona lity. MERC4 simulates methylation and demethylation reactions with a general equa tion for first-order kinetics (Martin, 1992): ] [ ] [20 R temp spec PC K f f t C (Eq. 7-1) CP is the concentration (mg/L) of product within a given mo del segment, free ionic (also referred to as dissolved or soluble) MeHg for methyla tion or free ionic Hg(II) for demethylation. CR is the concentration of reactant (m g/L) within a given model segment, free ionic Hg(II) for methylation or free io nic MeHg for demethylation. The rate multiplier, fspec, refers to the fraction of the tota l concentration of a mercury species (MeHg or Hg(II)) which is available for the r eaction, the dissolved fraction in this case. Free ionic (~soluble) concentrations are cons idered to represent the bioavailable fraction of total mercury, with soluble fraction define d as the fraction able to pass through a 0.45 m sieve. Dissolved fractions are based on measured partitioning constants (Bonzongo, 1996b) between dissolved and adsorbed phases, as indicated in Figure 4-2. Adsorption is a surface phenomena and is proportional to th e amount of surface area available (Carroll, 1999). Given equal volumes of particles of differing diameter, the total surface area becomes greater the smaller the pa rticle size; the rate of adsorp tion is indirectly related to the diameter of the adsorben t particles, an observation supported by data collected by

PAGE 160

146 Miller et al. (1999). K20 values are the reaction rates (day-1) as calculated from Carson River sediment samples (Carroll, 1999). To co rrect for water temperatures deviating from 20C, the temperature multiplier ftemp is used, as governed by the following equation (Martin, 1992), where T is the ambient temperature in C and Q10 values are constants calculated from Carson River sedi ment samples (Carroll, 1999): 10 20 10 T tempQ f (Eq. 7-2) Prior to the incorporation of bank elemen ts within the IFMTFM, methylation and demethylation only occurred in channel bed sediments. Sediment data suggests that methylation and demethylation rates in bed sediments vary with respect to spatial location, as indicated in Tabl e 7-1 (Carroll, 1999; Carroll et al., 2000). It is assumed that the chemical parameters that define each sample (pH being the only known parameter of significance) are not time dependant, so that no geochemical multipliers were used (Carroll and Warwick, 2001). In this study, the quotient of K20 (methylation) and K20 (demethylation) are referred to as the me thylation / demethylat ion ratio (M / D). Table 7-1. Spatial variability of M/D reaction constants within IFMTFM Range Segment Range,MethylationDemethylation (distance from CCG, km)Bed / Overlying Water Column K20 (day-1)Q10K20 (day-1)Q100.00 79.25 274 448 / 1 1750.00412.030.44832.32 79.25 95.40449 513 / 176 2400.00112.031.32912.32 95.40 115.00514 546 / 241 2730.00282.031.17532.32 The mass transfer of mercury from the channel bed sediments into the water column is calculated within the WASP5 code with equation 7-3, similar to Fick’s law (Carroll and Warwick, 2001): (Eq. 7–3) i i Di j j Di ij ij ij ij ij in C f n C f n L n A E t M /

PAGE 161

147 Equation parameters are defined as follows: M is mass (mg), t is time (days), A is interfacial area (m2), C is the total chemical concentration (mg/L), f is the dissolved portion of the chemical available for diffusion, L is the characteristic mixing length (0.01 m; Thomann and Mueller, 1987), n is porosity, E is the diffusion coefficient (m2/s), i is the ‘to’ segment, and j is the ‘from’ se gment. The surface area (A ) is dependant on the cross-sectional geometry and the hydraulic characteristics calculated in RIVMOD. The dissolved fraction of the chemical and concen tration gradient for diffusion are determined in MERC4 based on the first-order kinetic reactio ns for methylation and demethylation. The ‘old’ IFMTFM model is cal ibrated on three parameters for mercury, in order to best-fit limited water column samples for MeHg and inorganic Hg(II). Recall the assumed calibration relationships for inorga nic Hg(II) and MeHg within channel bank sediments, as shown in equations 4-1 and 4-2 in chapter 4 and earlier in this chapter. Mercury concentrations within eroded ba nk sediments remain constant over time, although concentrations are spa tially variable based on the 1 and 2 calibration parameters. The calibrated methylmercur y function predicted methylmercury bank concentrations over two orders of magnitude higher than the methylmercury bank data collected under conditio ns of near-zero flow (Carroll, 1999). The samples, from October 1994, were collected high up on the bank at the estimated point of ri ver contact during spring floods. Examination of water column data shows that, unlike other systems, increased discharge does not always cause a dilution in MeHg. Instead, an increase from low to medium flow causes a significant incr ease in MeHg water column concentrations. These results strongly suggest that given medium flows, ba nk erosion, and not diffusion,

PAGE 162

148 acts as the principal pathway for MeHg into the water column. This trend of MeHg increase with increasing flow begins to re verse when flows go from medium to high conditions. The resulting diluti on at higher flows suggests that the availability of MeHg from erosion of the upper bank is limited. An additional calibration parameter, a threshold depth (HT) above which MeHg bank concentra tion are taken as zero, was used to account for this observed dilution. Underl ying this assumption was the idea that high flow rates are too infrequent and short in duration to allow anaerobic microbes (performing the methylation) to develop in the upper portions of the channel banks. This same limitation is not observed for inor ganic mercury (Carroll, 1999). Modeling MeHg concentrations as constant al ong the vertical banks and over tim e is not representative of field conditions, which depend on microbe popula tions and therefore moisture levels. The IFMTFM includes three additional calibration parameters for the erosion and sediment transport functions. These parameters, and remain unchanged throughout this research and ar e not further discussed. Detailed information regarding these parameters is available from Carroll et al. (2002). Mercury Modeling within ‘New’ IFMTFM The functionality of the ne w bank elements within the IFMTFM is similar to the bed elements. The ‘new’ IFMTFM includes both bed elements and bank elements, whereas only bed elements were included in the ‘old’ IFMTFM. Methylation and demethylation occur within bank elements, resulting in temporally-variable MeHg concentrations not present in the ‘old’ IF MTFM, in which bank MeHg was only allowed to vary with respect to spa tial location according to the 2 calibration parameter. The parameters used to define bed elements remain unchanged from the ‘old’ IFMTFM, and

PAGE 163

149 bed MeHg concentrations remain dynamic. Di ffusion from contacted bank elements to the water column is modeled using the same parameter values and diffusion equation as the bed elements (equation 7-3). Bank elements differ from bed elements because bank elements possess two additional features. Bank elements can erode introducing mercury-laden sediment to the water column. This erosion process is cont ained within the ‘old’ model as well; the erosion functions and parameter values ( remain unchanged from those used in the ‘old’ model. Within the ‘new’ version, however, mercury concentrations adsorbed to erodable sediment and present within bank pore water are not constant, but rather continuously updated based on methylation and demethylation within bank sediments and equilibrium phase portioning (dissolved and ad sorbed). Advection is also modeled as a potential means for mercury transport fr om bank elements to the water column. Advection is not considered for bed segments. Bank element volumes are dynamic and equiva lent to the saturated portion of the channel bank. Recall the bank elements repr esent the vertical por tion of the assumed channel cross-sectional geometry, above the low-medium transition slope in Figure 3-1. The saturated heights of bank elements are a ssumed equal to the 4-day moving average of the water surface elevation within the adjacent water column elements (Figure 7-3). Four days is the length of time required for met hylating microbial populat ions to establish themselves and attain maximum methylat ion rates (Dr. Mark Hines, personal communication). Figure 7-3 show s a typical RIVMOD cross-sect ion; note that the water surface elevation, as predicted by RIVMOD, at the current model time step is likely not equal to the 4-day moving average water su rface elevation. If st reamflow (and hence

PAGE 164

150 river stage) is continually rising over the 4-day averag ing window, the average water surface elevation will be lower th an the current river stage; if streamflow (and hence river stage) is continually falling over the 4-da y averaging window, the average water surface elevation will be higher than the current rive r stage. For a given model segment, between two adjacent stations, the saturated height is computed as the arithmetic average of the saturated heights at the sta tions on either end of the segment. Due to symmetry, bank volume (saturated portion of bank element) is computed as the product of this average saturated height, the segment length (distan ce between stations), and the assumed active depth into the channel bank of 10 cm (equal to active depth for bed elements), multiplied by a factor of two (left and right bank). Figure 7-3. Relationship between tem porally-averaged RIVMOD stage and bank saturation height Predicted river stage for each model segment pot entially changes with each 60-second IFMTFM time step. The hydrodynami c (RIVMOD) portion of the IFMTFM is driven by daily flow values easily obtained fr om the USGS gages at Carson City, Fort Churchill, Truckee Canal, and Lahontan Dam. Mean daily flow values are assigned at time midnight for each day. Streamflow values do not remain constant throughout the ensuing 24-hour period, but rather are lin early interpolated from the values for consecutive days at each model time step. This approach ensures that streamflow varies 4-Day Average Water Surface Elevation at Current Model Time Step Saturated Height of Bank Element for Current Model Time Step

PAGE 165

151 gradually throughout the model domain. With flows updated for each time step, RIVMOD-predicted river stage also change s with each time step. The 4-day moving average for water surface elevation also th erefore changes with each model time step. Bank segment volumes (saturated portion of channel banks), howev er, are updated only at hourly intervals using this 4-day moving average, in the interest of minimizing computation time without a loss of accuracy. Within the saturated portion of a given channel bank segment, methylation and demethylation occurs at a pres cribed rate, based on model calibration to be discussed. The portion of the channel bank contacted by the water column is subject to loss of mercury through diffusion and erosion pro cesses. When bank element volumes are updated hourly, they are assumed to achieve instantaneous equilibri um with regard to volume and methylation rate. During a recedi ng hydrograph, river stage is falling and bank element volume is decreasing. At this in stant, the concentration of both total and soluble MeHg and Hg(II) in the channel bank segment will not change. However, the soluble mercury present in the now-dry porti on of the bank segment (product of change in bank segment volume, MeHg or Hg(II) concentration, and fraction of total Hg that is soluble) must go somewhere if solute continuity is to be preserved. This soluble quantity of MeHg and Hg(2+) is therefore introduced to the adjoining water column segment through advection. In the inte rest of completeness, advection and diffusion from the channel banks are modeled. However, it is noted that erosion remains the dominant mechanism for the introduction of mercury to the water column. Methylation within saturated channel bank elements requires a source of inorganic Hg(II) within the channel banks. As in the ‘old’ version of the IFMTFM code, the

PAGE 166

152 concentration of Hg(2+) within the channel ba nks is assumed to be temporally constant and spatially-variable between segment, based on calibration to the model parameter 1: (Eq. 4-2) The value of 1 does not change from the ‘old’ vers ion to the ‘new’ version. Model calibration is performed on the inorganic Hg (II) data set collected on May 16, 1994 under medium flow conditions of 600 cfs (27 cms). Model calibration is performed to achieve a simple visual best-fit through the limited num ber of sample data points. The calibration for 1 is shown in Figure 7-4, with a calibrated value of 1 = 1,900 g-Hg2+ / Kg-Soil. Model verification is shown in Figure 7-5 for high flow conditions (1,960 cfs, 55.5 cms) on June 10, 1995 and in Figure 7-6 for low flow conditions (62 cfs, 1.76 cms) on June 16, 1994. In these three figures, th e curve fit through the data poi nts is the same for the ‘old’ and ‘new’ versions of the code, as no modi fications were made with regard to the simulation of inorganic Hg(II). MeHg data is available for these same three data sets and is later used for MeHg calibration and verifi cation of the ‘new’ IF MTFM. Note that all concentration data for model calibration and verification are mercury concentrations in the water column. For reader reference, the three farthest downstream data points in the high and low flow data sets; and the two farthe st downstream in the me dium flow data set were collected in Lahontan Reservoir at th e river delta, the narro ws, and the dam. With a constant source of total inorganic Hg(II) within the channel banks, the soluble fraction of Hg(II) is determined based on the sediment composition and the equilibrium partitioning coefficients for the coarse suspended sediment (diameter > 0.063 mm) and washload (diameter < 0.063 mm) partic le sizes. The soluble fraction of Hg(II) is 5 0 0 1] [ S Hgbank in

PAGE 167

153 considered as the bioavailable fraction for methylation by sulfate-reducing bacteria. As previously discussed, methylation and demet hylation are modeled according to the first-order kinetic equation: (Eq. 7-1) Methylmercury is created within the saturated channel bank elements. With a higher fraction of fine sediment within the ch annel banks as compared to the channel bed, the reaction rates are not assumed to be the same. Instead, the K20 reaction rate for methylation is used as a new calibration para meter. Since methylation and demethylation reactions occur simultaneously, the model is calibrated on the net methylation rate, or rather the methylation / deme thylation ratio (M / D = K20 (methylation) / K20 (demethylation). With this in mind, the demet hylation rate within channel banks is taken 10 100 1000 10000 020406080100120 Distance from CCG (km)Inorganic Hg(II) Concentration in Water Column (ng/L) Measured Old & New Version1 = 1900 g/Kg Figure 7-4. Medium flow (May 16, 1994) IFMTFM calibration for inorganic Hg(II) (parameter 1) in channel banks for ‘old’ and ‘new’ code versions ] [ ] [20 R temp spec PC K f f t C

PAGE 168

154 10 100 1000 10000 020406080100120 Distance from CCG (km)Inorganic Hg(II) Concentration in Water Column (ng/L) Measured Old & New Version Figure 7-5. High flow (June 10, 1995) IFMTFM verification for inorganic Hg(II) (parameter 1) in channel banks for ‘old ’ and ‘new’ code versions 1 10 100 1000 10000 020406080100120 Distance from CCG (km)Inorganic Hg(II) Concentration in Water Column (ng/L) Measured Old & New Version Figure 7-6. Low flow (J une 16, 1994) IFMTFM verification for inorganic Hg(II) (parameter 1) in channel banks for ‘old ’ and ‘new’ code versions

PAGE 169

155 as equal to the demethylation rate within be d sediments, and the bank methylation rate is varied and calibrated on. In di scussing the required changes to the WASP input data file for the ‘new’ version of the IFMTFM code, it was mentioned that input data limitations allow a maximum of four separate methyla tion / demethylation rates based on segment classifications. In the ‘old’ input file, the me thylation and demethyla tion rates varied as indicated in Table 7-2. Note that methylati on and demethylation are not simulated in the water column, as conditions are no sufficiently anoxic. Table 7-3 indicates the rate coefficients us ed in the ‘new’ version. The addition of bank segments to the IFMTFM and the allo wance for methylation and demethylation within these segments requires specification of the reaction rate constants within bank sediments. Specification of K20 rates for bank segments in the ‘new’ code requires the removal of one of the previous four K20 values; K20 values for methylation and demethylation within the reser voir are therefore taken as the average of values previously specified for the upstream and downstream reser voir segments. As indicated in Table 7-3, the demethylation rate is held constant at the value used for river bed segments, and the bank methylation rate is calibrated to matc h observed water column samples for MeHg. The calibrated value for the methylation K20 rate is 0.014 day-1. This value is roughly three times greater than methylation rates observed in bed sediments, although it is noted that the demethylation rate was held constant at the valu e used for river bed segments. The net methylation rate, as repres ented by the M / D ratio is greater for bank sediments than bed sediments according to th is calibration. This observation can be explained, at least in part, by the larger per centage of fine sediment within the channel banks, as compared to the channel bed. Ba sed on field data (Heim and Warwick, 1997),

PAGE 170

156 Table 7-2. Segment classification for M/D reaction constants in ‘old’ IFMTFM MethylationDemethylation WASP Element TypeModel Segment Range K20 (day-1)Q10K20 (day-1)Q10Water Column1 2730.00000.000.00000.00 River Bed274 4480.00412.030.44832.32 Upstream Reservoir449 5130.00112.031.32912.32 Downstream Reservoir514 5460.00282.031.17532.32 Table 7-3. Segment classification for M/D reaction constants in ‘new’ IFMTFM MethylationDemethylation WASP Element TypeModel Segment Range K20 (day-1)Q10K20 (day-1)Q10Water Column1 2730.00000.000.00000.00 River Bed274 4480.00412.030.44832.32 Reservoir Bed449 5460.00202.031.25222.32 Channel Banks547 8190.014*2.030.44832.32 calibration parameter the following sediment descriptions are used in the IFMTFM model: river bed segments are modeled as 50% fine (diameter < 0.063 mm) and 50% coarse (diameter > 0.063 mm); river bank segments are modeled as 70% fi ne and 30% coarse; and reservoir bed and bank segments are modeled as 40% fine and 60% coarse (banks no longer distinguishable from bed). For Carson River sediments, the la rger surface area of fi ne sediments results in increased mercury sorption (Miller et al ., 1999), as indicated by the larger values shown in Figure 4-2. This higher sorption kine tic for fine particles results in a lower dissolved fraction of total mercury. Despite total mercury concentrations in the channel banks (higher percentage of fine sediment) t ypically one order of magnitude larger than along the channel bed, the disso lved fraction, or that frac tion assumed bioavailable for methylation and demethylation re actions, is lower than in th e channel bed sediments. In order to generate the amount of MeHg in the channel banks that is needed to match the water column observations, a higher methylation rate is therefore needed for input to the model. Although supporting data is not available, the calibrati on results are able to be justified. Also, the same result could have been achieved by lowering the demethylation rate.

PAGE 171

157 Comparison of ‘Old’ and ‘New’ IFMTFM Code for MeHg Modeling The ‘old’ version of the IFMTFM code required calibration on three parameters: 1 for inorganic Hg(II) in the channel banks, 2 for MeHg in the channel banks, and the threshold depth (HT) for MeHg in the channel banks. By creating the bank elements within the ‘new’ IFMTFM code, methylmerc ury concentrations in channel banks are actively simulated and continuously updated by the model. Rather than calibrate on 2 and HT, the model is only calibrated on the K20 rate for methylation in the channel banks. For the ‘new’ code, the model is calibrated using the medium flow data set (May 1994; Q = 600 cfs, 27 cms), as shown in Figure 77. Model verification uses the high (June 1995; Q = 1,960 cfs, 55.5 cms) and low (June 1994; Q = 62 cfs, 1.76 cms) flow data sets, as shown in Figures 7-8 and 7-9. Before th e code modifications, the medium flow data set was used for 2 calibration ( 2 = 4.57 g-MeHg / Kg-soil), and the high flow data set was used for HT calibration (HT = 1.00 m). To facilitate comparison, IFMTFM results for ‘old’ and ‘new’ versions of the code are shown in each figure. The environmental concerns regarding methylmercury underscore the importance of accurately modeling this species within the Carson Ri ver IFMTFM model. Channel banks have higher concentrations of mercury associated with sediment than those found within the channel bed (Miller and Lechler, 1998). Mercury is, of course, not uniformly distributed throughout the system; instead, bank concentrations vary over three orders of magnitude with localized “hot spots of c ontamination” (Miller and Lechler, 1998).This thesis improves the ability of the IFMTFM to simulate processes which control MeHg concentrations within channel banks. Bank erosio n is then a controlli ng factor for water

PAGE 172

158 0.0 2.0 4.0 6.0 8.0 10.0 12.0 020406080100120 Distance from CCG (km)MeHg Concentration in Water Column (ng/L) Measured Old Version New Version Figure 7-7. Medium flow (May 16, 1994) IFMTFM calibration for MeHg ( 2) in channel banks (‘old’ code version) and K20 methylation rate in channel banks (‘new’ code version) 0.0 2.0 4.0 6.0 8.0 10.0 12.0 020406080100120 Distance from CCG (km)MeHg Concentration in Water Column (ng/L) Measured Old Version New Version Figure 7-8. High flow (June 10, 1995) IFMT FM calibration for MeHg threshold height (HT) in channel banks (‘old code version) and verification for MeHg in channel banks (‘new’ code version)

PAGE 173

159 0.0 2.0 4.0 6.0 8.0 10.0 12.0 020406080100120 Distance from CCG (km)MeHg Concentration in Water Column (ng/L) Measured Old Version New Version Figure 7-9. Low flow (June 16, 1994) IFMTFM verification for MeHg in channel banks for ‘old’ and ‘new’ code versions column concentrations of merc ury because of the high affinity of mercury for particulate matter. Comparing the calibration and verification of the ‘old’ and ‘new’ code versions for water column MeHg concentrations, the ‘new ’ code is clearly superior for matching observed concentrations of MeHg within La hontan Reservoir, based on visual inspection of the respective model output for distances greater than 85.74 km (model segment 201) downstream of CCG. The ability of the IFMT FM to duplicate these reservoir data points is particularly important given the bio accumulation processes o ccurring within the reservoir. With respect to data points betw een CCG and the reservoir, matching of data points between the two versions is comparable for the medium flow data set (Figure 7-7), improved by the ‘new’ version for the low flow data set (Figure 7-8), and diminished by the ‘new’code for the high flow data set (F igure 7-9). The usefulness of a quantitative

PAGE 174

160 comparison is limited by the small number of data points to which the calibration is being fit. It is additionally noted that the MeHg c oncentration peak at kilometer 85 in Figure 7-9 is not a model error, but rather the result of three f actors: high mercury concentrations in bed and bank sediments, based on equati on 4-2 and the small channel slope in the deltaic region; high reservoir stage, associat ed with previous high flows, and reservoir contact with bank sediments caused by this backwater effect; and high opportunity for diffusion of MeHg into the overlying water column because of low flow velocity. Measurement error associated with the me rcury sample points does not impact the perceived fit between sample data points and the IFMTFM model. Error bars for the total MeHg sample data were determined from water column samples taken on June 16, 1994 (Table 7-4, points in Figure 7-9). Samples we re taken in triplicate at six locations along the Carson River mainstem (including the de lta) and two locations within Lahontan Reservoir. The average error (standard deviati on) for samples collected along the river is 0.056 ng/L, a value not discernable when applied to sample data points in Figures 7-7, 7-8, or 7-9. Only the data point collected at Fisherman’s Point within the reservoir demonstrates a high sample standard devi ation, likely the resu lt of high sediment concentrations (poor sample uniformity) associ ated with a previous period of high flow. A qualitative comparison has been considered, but more important is the ability of the ‘new’ code to better simulate natural processes occurring in the field. During the upgrade from ‘old’ version to ‘new’ version, no change s were made to sediment composition (in terms of percent fines and percent coarse, within channel banks or channel bed), which determine the fraction of adsorbed versus soluble and bioavailable mercury. Additionally, no changes were made to the erosion functions or values of

PAGE 175

161 Table 7-4. Error associated with sample MeHg data points Distance from Mean Total MeHg*Standard deviation* CCG (km)(ng/L)(ng/L) 00.5120.054 110.5050.067 281.6070.087 442.0010.074 691.7860.031 871.8130.023 953.7251.090 1110.4000.033 computed from 3 samples at each location calibration parameters for erosion. IFMTFM parameters which remained unchanged between ‘old’ and ‘new’ code versions are summarized in Table 7-5. The modeling of MeHg within channel banks of the Carson Ri ver was changed by this thesis research. A comparison between the ‘old’ and ‘new’ versio ns of the IFMTFM, as related to modeling of bank MeHg, is shown in Table 7-6. Table 7-5. Parameters unchanged betw een ‘old’ and ‘new’ IFMTFM versions Channel bed:no erosion (vertical equilibrium) dynamic MeHg concentrations Hg(II), MeHg diffusion to water column Channel banks:lateral bank erosion Hg(II) constant over time Hg(II) varies spatially: [Hgin] = 1 / S0 0 5 General:sediment composition unchanged sediment transport parameters unchanged partitioning coefficients unchanged Table 7-6. Comparison of channel bank MeHg modeling in ‘old’ and ‘new’ IFMTFM old' versionnew' version MeHg temporally constantMeHg varies temporally with dynamic zone of active methylation MeHg varies spatially: [HgMeHg] = 2 / S0 0 .5MeHg varies spatially based on changing river stage and bank contact threshold depth (HT) for MeHg no threshold depth for MeHg no diffusion of Hg to water columndiffusion from contacted banks no advection of Hg water columnadvection during receding hydrographs calibration on 2 and HTcalibration on bank K20 methylation rate MeHg available to water columnupon bankMeHg available following 4-day contact microbial acclimation time

PAGE 176

162 The ‘old’ version of the IFMTFM code wa s calibrated to MeHg concentrations in the bank, which were allowed to vary between segments but not over time. Given that the methylation and demethylation reactions are microbially-mediated, MeHg concentrations cannot be static through time. Mercury methyl ation is stimulated if soils are fully saturated and anoxic conditions are achieved (Gilmour and Henry, 1991; Porvari and Verta, 1994). High flow rates ar e too infrequent and short in duration to allow anaerobic microbes to actively methylate mercury in the upper portions of the bank sediments throughout the year. This said, there is not however, a threshold height above which MeHg within channel banks will always be ze ro, as assumed by the ‘old’ code structure. The implications of this change are seen in Figure 7-10, a flood simulation for May and June of 1995 during which all flows contacte d the banks above the 1.00-meter threshold depth. In the ‘old’ version of the code, th e threshold height for MeHg introduction into the water column results in a dilution effect (no additional MeHg available as flows rise). The ‘new’ code version allows MeHg concentr ations to be created and eroded from the upper channel banks; additionally, the time lag for microbial response (as captured by the use of the 4-day averaging window for the height of bank saturation) is evident compared to the flow hydrograph. Sample data for MeHg is not available for this time period. For comparison, a simulation from May and June of 1994 is also included as Figure 7-11. Here, all flows have a depth less than the 1.00-meter threshold depth. The ‘old’ code version has MeHg immedi ately available for input to the water column as flow rises. The ‘new’ code shows a delay, following flow rising, in creating the MeHg within the channel banks and subseque nt input into the channel. MeHg concentrations within contacted portions of the channel banks, previously m odeled as susceptible only to

PAGE 177

163 0 2 4 6 8 10 1205/13/9505/20/9505/27/9506/03/9506/10/95MeHg Concentration in Water Column (ng/L)0 30 60 90 120Flow (cms) New Version Old Version Flow Figure 7-10. IFMTFM simulation for May / June 1995 flood event at Fort Churchill (model segment 140), illustrating MeHg d ilution effect of threshold height HT in ‘old’ code version 0 2 4 6 8 1005/19/9405/26/9406/02/9406/09/9406/16/94MeHg Concentration in Water Column (ng/L)0 2 4 6 8 10Flow (cms) New Version Old Version Flow Figure 7-11. IFMTFM simulation for May / June 1994 flows at Fort Churchill (model segment 140), demonstrating time lag for MeHg response in ‘new’ code version

PAGE 178

164 erosion, are also not temporal ly constant in the ‘new’ code In this regard, the ‘new’ IFMTFM code is superior. Bank element volum es are dynamic dependent on the time-history of river stage. Methylation and demethylation processes actively occur within saturated portions of the channel ba nks, imparting dynamic MeHg concentrations within channel banks. MeHg concentrations now display a dependence on soil moisture conditions, as evidenced by prev ious field data (F igure 4-3), within both channel bed and channel bank sediments. Contacted portions of the channel banks are no longer subject onl y to erosion, but the ‘new’ code also allows for mercury introducti on to the water column through diffusion; advection of mercury to the wate r column is modeled during periods of a receding hydrograph. Though advection and diff usion are not the primary contributors to water column mercury concentrations (ero sion), the processes no doubt do occur in the field. In addition to improving the model capab ilities to simulate field processes, the ‘new’ code reduces the number of calibration parameters for mercury from three to two and provides an improved match to in-channel MeHg observations. The Next Step for the IFMTFM Model While this thesis research has addres sed one IFMTFM shortcoming by adding the ability of the model to simu late dynamic methylmercury con centrations within channel banks, mercury modeling within the IFMTFM continues to have room for improvement. Two specific opportunities will be briefly discussed. Bioavailable mercury for methylation a nd demethylation reactions is presently taken as equivalent to the so luble (or dissolved) fraction of total mercury. The soluble fraction is computed within MERC4 through th e use of calculated pa rtition coefficients (Kd values). Partitioning coefficients were measured only for the washload fraction of

PAGE 179

165 sediment (the fine sediment) and were co mputed for the coarse suspended solid and bedload fractions by making use of average pa rticle diameters and the washload results (Bonzongo, 1996a). From the original “grab” sample collected just below the water surface of the Carson River, the washload fraction and dissolved fractions were functionally defined as the samp le fractions captured by a 0.45 m sieve and passing through the sieve, respectively. The partitioning coefficient (L / mol) was calculated for MeHg and Hg(II) using equation 7-4, where [HgP] is the concentration of mercury bound to washload particles in the sample, [HgD] is the concentration of dissolved mercury in the sample, and [Solid] is the concentration of the solid constituent in molar units, assuming the solids are composed of iron hydroxide (FeOOH): Solid Hg Hg KD P d (Eq. 7-4) A sediment density of 1.9 kg/L is used by the model, as computed from a measured sediment dry bulk density of 1.6 g/cm3 (Miller et al., 1999) and assuming a porosity of 0.30 (Carroll, 1999). The IFMTFM predicts th e dissolved and adsorbed fractions of MeHg and Hg(II) using this sediment characteristic and the computed Kd values. These fractions are then applied to the total me rcury concentrations within bed and bank segments to determine the amount of solubl e MeHg available for methylation and the amount of soluble Hg(II) available for deme thylation. Methylation and demethylation kinetics then govern the dynamic concentra tions of both species within bed and bank segments throughout the model domain. The mo deling approach therefore makes use of the assumption that soluble me thylmercury and soluble inor ganic mercury is the only fraction of total mercury available to th e methylating and demethylating microbial populations. More research is ongoing to de termine the validity of this assumption.

PAGE 180

166 Reasonably, it can be assumed that the microbes are not floating in solution but rather are themselves attached to the sediment grains, in which case they may be able to make use of some portion of adsorbed mercury as well. Again, this may explain the need to use a significantly higher met hylation rate in bank elements. After the incorporation of bank elements into the IFMTFM, one other problem also emerged. This is the problem of limited data. Data available for inorganic mercury concentrations within the ch annel banks along the Carson Ri ver are highly variable for multiple samples at a given site and spatially along the river (Miller and Lechler, 1998). The current IFMTFM version is calibrated to the methylation rate in the channel banks, and the model predicts spatially-variable and time-dependant MeHg concentrations. However, no sample data is available to verify these model predictions. To facilitate such a comparison, additional data on the distributi on of mercury in channel banks is needed. While the model predictions are in good agr eement with water column data for MeHg and Hg(II), the question remains whether th e model predictions are good for the right reasons. Bank elements each possess unique Hg (II) concentrations, but doubts remain as to whether the 1 calibration parameter is the best modeling approach to capture the spatial variability.

PAGE 181

167 CHAPTER 8 COMPARATIVE ANALYSIS: RIVMOD VERSUS FESWMS MODULE OF SMS When the SMS model was selected as the surface water model for this project, two distinct end-products were envisioned. The firs t was to develop a simple relationship, to relate stage in the channel to water table level within the bank. This thesis research has concluded that the water table response to stage changes is approxi mately instantaneous. Given this rapid response time, the assumption is made that the free surface elevation in the channel is equal to the phreatic surface in the channel bank immediately adjacent to the channel. The saturated por tion of the erodable face of the channel bank varies temporally with the river stag e, and the two water surface el evations are equivalent. To incorporate this relationship into the functionality of th e new IFMTFM bank elements, microbial response time for methylation and de methylation reactions was also necessarily considered. Laboratory data sugge st a time period of 4 days is needed for the methylating population to become established and attain peak reaction rates (Dr. Mark Hines, personal communication). Because the M/D rati o in the channel banks is a calibration parameter, rates for methylation and demethylation reaction rates (K20) values were not allowed to vary with time. Instead, the cal ibration parameter was taken as the peak M/D ratio, and the height of the saturated (activ ely methylating) portion of the channel banks is dynamically computed from the four-day moving average of the water surface elevation in the channel and the ch annel geometry used in RIVMOD. The second end-product is a comparative analysis between RIVMOD predictions and SMS predictions. For the modeled Scout Camp reach, a comparison between the two

PAGE 182

168 models was sought for flow depths and mean velocity to see if the simplified channel geometry and 1D nature of RIVMOD resulte d in consistently high or low predictions relative to a more detailed and calibrated m odel. The paucity of surveyed water surface elevation data prevented calibration of the SMS model. Limitations within the SMS model also mandated the assumption of a te mporally-constant head loss through the SMS model domain of 1.99 meters, irrespective of flow magnitude or antecedent flow conditions. Antecedent flow conditions are al so excluded based on the model’s inability to simulate element re-wetting (as would occur for rising portions of a dynamic streamflow hydrograph) without loss of numerical stab ility. The FESWMS-2DH model mesh created using SMS software is clearly no t the accurate predictor of river stage that was originally envisioned before the start of this thesis research. The accuracy of SMS stage predictions over time cannot be quantified, as only a si ngle data point is generated from surveyed water surface elevation data in the vicinity of the wells. Through regression analysis, shown in Figure 6-27, th e observed water surface elevation at the well transect was estimated as 1252.30 mete rs (NAVD 88) for the streamflow of 173.8 cfs on March 9, 2001. Comparing this field observation to the rating curve prediction from SMS model results, the SMS model pr ediction is 15 centimeters high, at an elevation of 1252.45 meters for the wate r surface. A comparison between SMS and observed water surface elevations in the channe l is not available for any other time during the 2001-2002 study of the Scout Camp reach of the Carson River. With these considerations in mind, the comparative an alysis will still be conducted, although the limitations of the analysis are noted in advance.

PAGE 183

169 Comparison of Channel Geomet ry along the Well Transect In general, RIVMOD is based on a generali zed channel shape (figure 3-1) and data interpolated from topographic maps, and FE SWMS-2DH (SMS) is based on precise data collected using GPS. A visual comparison of the channel cross-section (at the location of Well 1) between the two models is shown in Figure 8-1. The figure is plotted by normalizing all elevations with respect to th e minimum bed elevation along the respective transect. The SMS transect is extracted direc tly from the SMS mesh. The geometry of the main channel (below floodplain stage) in RIVMOD utilizes a simplified cross-sectional shape and average values for each geometric parameter (as indicated in Figure 3-1) and these parameters do not vary spatially. It is therefore not surprisi ng that the model cross-sections differ in shape. In particular the main channel width used in SMS (52.61 m) is greater than the channel width used by RIVMOD (45.50 m) by 7.11 meters, and the channel bank height in RIVMOD (3.27 m) is significantly larger than the average channel bank height in SMS (2.30 m). One potenti al opportunity for IFMTFM improvement would involve the collection of higher resolution cross-sectio n data and the incorporation of this data into RIVMOD segment geomet ry. The Manning roughness coefficient for the main channel is equal to 0.045 within both m odels. This value was selected to properly calibrate the 1D RIVMOD model, but it likely high for actual field conditions. No data is available regarding the setting of this parameter, and lower values within SMS are not tested based on time constraints a nd model run time requirements. Steady-State Comparisons for Velocity and Depths of Flow Model output for velocity and depth of flow were compared for the SMS and RIVMOD models over the applicable range of the SMS model. While RIVMOD includes the ability to simulate flows spilling onto the fl oodplain, this capability is not available to

PAGE 184

170 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 -40.00-30.00-20.00-10.000.0010.0020.0030.0040.00 distance from transect midpoint (m)elevation above thalweg (m) Well Transect in SMS Well Transect in RIVMOD 45.50 meters 52.61 meters 2.16 meters2.435 meters 3.27 meters Figure 8-1. Comparison of channel geometry at well transect for SMS and RIVMOD SMS because of the required inclusion of vertical wall model boundaries. At the well transect in SMS, channel banks on both sides of the channel are characterized by vertical walls (Figure 5-3). These walls extend upward indefi nitely, but the surveyed elevation for the top of the channel bank nearest the we lls (1254.15 meters,NAVD 88) is used to indicate the top of the cha nnel banks in Figure 8-1. Based on the rating curve analysis (Figure 6-34) and associated discussion found in chapter 6 of this thesis, recall that the applicable flow range for the SMS model is 1.91 cms (67.5 cfs) to 109.6 cms (3871 cfs). The rating curve equation was developed us ing 0.25-meter head increments along the upstream and downstream flow boun daries, resulting in 7 data points falling within the applicable flow range of SMS. The same se ven data points (discharge values) used to develop the rating curve equation for depth as a function of discharge will be used as the

PAGE 185

171 basis for the comparison of SMS and RIVMOD solutions for velocity and depth at the well transect (Table 8-1). Table 8-1. Comparison of steady-state ve locity and depth output from SMS and RIVMOD Discharge, cms109.684.263.745.320.815.26.85 (cfs)(3871)(2974)(2250)(1600)(735)(537)(242) SMS Average Velocity (m/s)1.130.980.870.760.380.400.35 RIVMOD Average Velocity (m/s)1.121.020.910.800.590.520.38 SMS Average Depth (m)1.891.671.421.171.070.740.39 RIVMOD Average Depth (m)2.532.211.921.631.161.030.78 SMS Contacted Bank Height (m)1.751.531.281.030.930.600.25 RIVMOD Contacted Bank Height (m)1.911.581.291.000.530.400.15 Bank Contact Difference (m)*0.160.050.01-0.03-0.40-0.20-0.09 The following procedures were used to ge nerate the values in Table 8-1. Both models are allowed to run to steady state befo re solutions are extracted. Noting that the well transect is approximately centrally-l ocated between RIVMOD stations 152 and 153 (segment 152), RIVMOD average velocity and RIVMOD average depth are computed simply by averaging the steady-state model out put for the given flow magnitudes at the two stations. The 1D RIVMOD model generates a single output value at each station for each time step. Flow depths are equivalent to the depth above the channel thalweg (elevation 0.00 meters in Fi gure 8-1). Contacted bank hei ght was computed as the difference between flow depth and the eleva tion at the bottom of the RIVMOD vertical banks (0.63 meters). The procedure di ffered for SMS output. The non-uniform node spacing along the well transect required this to be considered. Model output is available at each node for velocity magnitude and flow depth above the nodal bed elevation. Depth-averaged velocity and depth vary across the channel. A simple arithmetic average would be appropriate if node spacing were uniform. This not being the case, numerical integration via the trapezoidal rule was used to compute the average velocity and flow depth for all nodes along the transect. Alt hough all nodes do not lie perfectly on the well

PAGE 186

172 transect, this simplification is reasonably made, and node locations (and thus distance between nodes) were able to be computed with respect to their distance from the left channel bank, looking upstream. The contacte d bank height for SMS was computed as the difference between flow depth and the av erage elevation of the bottom of the SMS vertical banks in Figure 8-1 ( (0. 00 m + 0.275 m) / 2 = 0.1375 meters). Velocity comparisons are shown in Figure 8-2 and Figure 8-3 for discharge values of 109.6 cms (3871 cfs) and 6.85 cms (242 cf s), respectively. At the higher flow magnitude, average velocity computed usi ng SMS is greater than average velocity computed using RIVMOD, while the reverse is true for the lower flow magnitude. Velocity predictions between the two models differ by less than 0.05 meters per second for all but two of the seven comparisons, with significant differences of 0.12 m/s and 0.21 m/s at discharge values of 15.2 cms a nd 20.8 cms, respectively. Of the seven comparisons, RIVMOD velocity was exceeded by SMS velocity only at a discharge of 109.6 cms. The two models compare most favorably under high flow conditions (Q > 1000 cfs) most important for mercury modeling. Under these high flow regimes, bank erosion is the primary determinant fo r mercury introduction to the water column; bank erosion, as modeled with in the IFMTFM, is highly corre lated to velocity so that higher velocities result in ever-increasing erosion of mercury-contaminated bank material. The changing shape of the velocity di stribution (SMS node solutions) across the channel is also of note in th ese figures. This is the result of the bed morphology. High points along the bed present increased frictional resistance to flow, resu lting in alternative preferential flow pathways wh ere the channel is deeper. Tu rbulence and steep velocity

PAGE 187

173 0.00 10.00 20.00 30.00 40.00 50.00 60.00 0.000.200.400.600.801.001.201.40 velocity magnitude (m/s)distance along transect (m) SMS Node Solutions SMS Average RIVMOD Average Steady-State Flow through Well Transect = 109.6 cms A verage SMS Velocity = 1.13 m/s A verage RIVMOD Velocity = 1.12 m/s Figure 8-2. Q = 109.6 cms: comparison of steady-state velocity output from SMS and RIVMOD 0.00 10.00 20.00 30.00 40.00 50.00 60.00 0.000.200.400.600.801.001.201.40 velocity magnitude (m/s)distance along transect (m) SMS Node Solutions SMS Average RIVMOD Average Steady-State Flow through Well Transect = 6.85 cms A verage SMS Velocity= 0.35 m/s A verage RIVMOD Velocity = 0.39 m/s Figure 8-3. Q = 6.85 cms: comparison of steady-state velocity output from SMS and RIVMOD

PAGE 188

174 gradients associated with boundary layer e ffects are not considered in SMS model computations; vertical walls do cause frictiona l shear stress and force tangential velocity at these nodes, but velocity computations ex clude the boundary layer effects, and velocity is therefore not equal to zer o at these no-flow boundaries. Depth comparisons are shown in Figure 8-4 and Figure 8-5 for discharge values of 109.6 cms (3871 cfs) and 6.85 cms (242 cfs), respectively. For the seven compared discharge values, the average predicted depths by RIVMOD always exceed the average predicted depths from SMS. This is to be expected because of the lower channel width used in RIVMOD. Of more consequence is the comparison of the height of contacted channel bank between the two models, as this value will be used to predict the active methylation zone within the newly-incor porated bank elements of the IFMTFM. The contacted height of rive r stage on the vertical cha nnel banks varies between the SMS and RIVMOD models. The magnitude of this bank contact difference was previously indicated in Table 8-1. The di fference is more evident when the SMS and RIVMOD predictions are plotted with respect to a line of perfect agreement, as shown in Figure 8-6. If the two models were in perfect agreement, the predicted bank contact height would be equal for the two models fo r each discharge, and all data points would plot on the line of perfect agreement. Bank contact heights predicted using SMS, on average, exceed bank contact heights predicted using RIVMOD by 7.1 centimet ers. Error statistics may be computed using model results shown in Table 8-1, but these statistics are of little value given the uncertainty with SMS model output. The SMS model is not properly calibrated to water surface observations, and it was necessary to as sume a constant head loss across the

PAGE 189

175 Figure 8-4. Q = 109.6 cms: comparison of st eady-state depth predictions from SMS and RIVMOD 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 -30.00-20.00-10.000.0010.0020.0030.00 distance from transect midpoint (m)elevation above thalweg (m) Well Transect in SMS Well Transect in RIVMOD SMS Average Depth RIVMOD Average Depth RIVMOD Average Depth = 1.03 m SMS Average Depth = 0.74 m Q = 15.2 cms (537 cfs) Figure 8-5. Q = 6.85 cms: comparison of st eady-state depth predic tions from SMS and RIVMOD 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 -30.00-20.00-10.000.0010.0020.0030.00 distance from transect midpoint (m)elevation above thalweg (m) Well Transect in SMS Well Transect in RIVMOD SMS Average Depth RIVMOD Average Depth RIVMOD Average Depth = 2.53 m SMS Average Depth = 1.89 m Q = 109.6 cms (3871 cfs)

PAGE 190

176 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80 2.00 0.000.200.400.600.801.001.201.401.601.802.00 SMS Contacted Bank Height (m)RIVMOD Contacted Bank Height (m)Line of Perfect Agreement Figure 8-6. Comparison of c ontacted bank height predicti ons from SMS and RIVMOD model domain, regardless of flow magnitude. The availabil ity of additional calibration data would prove useful to th is type of error analysis, but the required data is not available for this thesis research. Limited data and limited capabilities of the SMS model prevent an accurate quantit ative comparison between the detailed SMS model (assembled from surveyed channel bathymetry) and the simplified RIVMOD model (utilizing simplified channel geometry and average para meter values). Recall, however, that the water surface elevation predicted usi ng SMS exceeds the surveyed water surface elevation from March 9, 2001 by 15 cm. As di scussed, the SMS model results include several assumptions and simplifications su ch that the SMS predictions represent maximum levels that could be observed in the channel. If additional data were available to correct the shortcomings of the SMS mode l, the two models would likely illustrate improved agreement under the high flow condition s at which bank contact is particularly

PAGE 191

177 important for determination of a methylati ng bank zone and erosion of contaminated bank material. Comparative Analysis and Data Limitations The comparative analysis chapter of this thesis serves as a microcosm of the difficulties encountered throughout this thesis research. Detailed spatial and temporal modeling efforts require extensive data for model calibration. Unfortunately, for this research, the SMS model presented unfores een limitations given the data which were available. Additional water surface elevations for different streamflow regimes would have been a valuable tool to assist in SMS model calibration and a comparison of stage predicted by SMS to water table observati ons and RIVMOD stage predictions. For rating curve development, measurements to validate or invalidate the accu racy of streamflows routed using RIVMOD would also have been useful. However, making use of available data, alternative analyses were able to quan tify the coupling of rive r stage and water table oscillations. The primary deliverable for this thesis wa s the incorporation of a new bank element into the IFMTFM, to relate flow history with in the channel to moisture content levels observed within the channel ba nks. Zones of varying moistu re impact methylation and demethylation kinetics, and therefore the te mporal concentrations of methylmercury available to the water column. Data analysis indicated a near ly instantaneous response of the water table to streamflow oscillations, at a monitoring well 5 f eet removed from the channel. The IFMTFM is only concerned with the erodable surface of the channel banks. It was assumed that for flow levels contacti ng the channel bank, the erodable face of the channel bank would be saturated up to a he ight equivalent to the river stage. To incorporate microbial acclimation time into the IFMTFM, bank element volumes vary

PAGE 192

178 over time as river stage levels change. The saturated portion of the channel bank elements are computed based on the 4-day moving aver age of RIVMOD-predicted river stage. Incorporation of the bank element into the complete Carson River model of hydrodynamics, sediment transport, and me rcury transport and cycling (IFMTFM), provides the model with a new capability to simulate temporally-variable mercury concentrations within the channel bank, re duces the model calibration procedure, and increases the accuracy of in-stream mercury predictions, while also adding to the model’s ability to simulate processes actually occurring in the field.

PAGE 193

179 APPENDIX A SMS MODEL PARAMETERIZATION AND MESH EDITING Creation of Triangular Mesh from Scatter Point Data Set After the set of surveyed data points are imported into SMS as scatter points within the scatter module, the model us er selects the coverage type under ‘Data / Switch Current Model’ to FESWMS and switches to the SMS river module to create and edit the mesh and parameterize the model. The options available within the SMS river module are determined by the selected coverage. In this case, FESWMS options are available and the FESWMS model is run from this module. The software incorporates numerous automatic mesh generation techniques, of which triangulation is most suitable fo r a network of random survey points. The triangulation generates a mesh of either quadr atic or linear triangles with the quadratic option creating additional nodes at the midpoint of each triangle leg. A single node corresponds to each point in the scatter data se t. It is possible to toggle between element types using the ‘Elements / Linear <-> Qu adratic’ feature of the FESWMS module. Despite the implied increase in computational effort, quadratic elements are selected in order to increase numerical stability during the wetting and drying process. Triangular elements are created from the scatter point set by selecting ‘Eleme nts / Triangulate.’ The triangulation process creates elements outsi de the desired mesh boundaries, a by-product of the non-linear shape of the channel. The mode l user is required to manually select and delete these extraneous elements. Thin triangles along the mesh boundary, the corner nodes of which are almost co-linear and eff ectively hidden, are identified according to a

PAGE 194

180 thin triangle aspect ratio (perfect equilatera l triangles have an asp ect ratio of 1.0) set under ‘Elements / Options.’ With the defau lt value of 0.04 unchange d, these triangular elements are selected under ‘Elements / Select Thin Triangles’ and deleted. Transformation of Mesh from Tria ngular to Quadrilateral Elements After the deletion of outer elements, the ge neral geometric and bathymetric features of the channel can be distinguished. At this po int it is useful to consider the desired mesh characteristics. The mesh must be able to simulate the areal wetting and drying of the model domain. Vertical channel banks result in a drying pattern from the floodplain to the outer channel banks to the inner cha nnel thalweg. The final model mesh should consist of elements that are approximately parallel to the channel bank boundaries. This will allow the model domain to dry uniformly; perhaps the best analogy is the peeling of an onion. Model functionality di ctates that as one node is pr edicted to be dry (predicted water surface below node elevation), the enti re element becomes dry and incapable of transporting flow. Failure of the elements to fo llow this basic design will result in jagged boundaries of the water edge, isolated ponded areas, and resulting numerical divergence. Quadratic elements are best able to captur e this wetting / drying fr ont and offer improved numerical stability and a faster solution time for the model (SMS Tutorial: Version 8.0, Section 3.4). Triangular elemen ts will be necessary is some areas to improve model wetting and drying in steep area and around bends in the channel. At this point, the mesh is composed entirely of triangles. Quadrilateral elements are preferred because of the need to simula te element wetting and drying. Triangular elements are converted to quadrilateral elem ents using the ‘Element / Merge Triangles’ feature in FESWMS. Before merging triangles the merge triangles feature angle is set under ‘Elements / Options.’ Using the SMS mode l default, triangles are only merged if

PAGE 195

181 all angles of the resulting qua drilateral are greater than 65 SMS merges triangles in an iterative manner, starting w ith quadrilateral angles of 90 and working down in a step fashion, to form the best-shaped elements. Th e finite element method is more stable and accurate when quadrilateral elements are rectangular and tria ngular elements are equilateral (SMS Tutorial: Version 8.0, Lesson 3.5). Clearly, a mesh cannot consist entirely of these ideal shapes, but the elemen ts should approach these shapes as close as possible. To toggle between 8-noded and 9-noded ( nodes at all side midpoints and in the center) quadrilateral elements, use the option under ‘Elements / QUAD8 <-> QUAD9.’ 9-noded quadrilateral elements decrease node spac ing and should assist the finite element scheme with model convergence. For FESWMS it is best to use 9-noded quadrilateral elements, even though both 8-noded and 9noded quadrilaterals are supported (SMS Tutorial: Version 8.0s, Lesson 5.2). The model mesh after merging of triangles is shown in Figure A-1. With a basic mesh in ha nd and knowledge of the desired end product, mesh shortcomings are next addressed and remedied. Mesh Corrections Based on Perceived Model Operation The first round of mesh corrections is pe rformed in advance of using the finiteelement model, or, alternatively, with no water flowing in the channel. The corrections are based solely on observed n eeds and intuition with regard to how the model performs calculations. As mentioned previ ously, an abundance of points is available at each of the five cross sections. Points along the top of banks and on the floodplain are incomplete or lacking in some locations. Points at the base of vertical banks are also very limited, due to limitations associated with the surveying equi pment. Lastly, channel bed features, most

PAGE 196

182 notably the thalweg, are poorly represented. All of these elements are critical if water is to actually flow through the model domain and channel stage is to be accurately simulated. Figure A-1. Quadrilateral element mesh af ter initial merging triangular elements Rectifying the Mesh with Observed Channel Geometry The FESWMS module allows the user to manually create mesh nodes by clicking on a desired location. Within the node opti ons dialog box, check the option to insert nodes into the triangulated mesh and retriangu late voids when deleting. By using these options, inserted nodes are automatically upda ted with elevations to conform to the existing elevation contours, and nodes deleted to improve the node sp acing will not result in gaps in the mesh. One other model feat ure of note when creating new nodes is the interpolation option. This allows the user to select two nodes of known elevations and generate any number of additional nodes between with spatial position and elevation determined by linear interpolation. elevation 1250.30 1250.80 1251.30 1251.80 1252.30 1252.80 1253.30 1253.80 1254.30 1254.80

PAGE 197

183 Admittedly, the addition and deletion of node s cannot be performed arbitrarily. The goal is to develop a mesh that accurately po rtrays the observations of channel geometry within the model domain. Were this not the case, the final model predictions of river stage, to be compared against water tabl e stage, would be of minimal validity. In general, water flows downhill. This is not a scientific breakthrough, but it is an important starting point for re fining this model mesh. Some basic limitations to numerical modeling also need to be realized. For a given simulation, three solution values are generated for each node: water surface elev ation, water depth (a function of node elevation and water surface elevation), and a ve locity vector. The equations built into the SMS solution algorithm were previously re viewed, but these pr ovide little insight compared to simple experience with usi ng this model. The knowledge obtained from model reference manuals, model tutorials, and simple trial-and-error experimentation was used extensively during this stage of mode l development. With numerical modeling, a divergent solution at a single node will cause the model to crash. The important point taken from all of this is that bed elevation needs to be gradually lowered downstream in the direction of flow. Flow pathways need to be determined in advance of mesh refinement. Information from first-hand observations, photographs, names given to survey points (such as thalweg, water surface, or bank), and graphical analysis of survey data were used to discern the channel thalwe g, connect thalweg points to each other, and verify that the general characteristics of th e channel bed (high points and low points) are representative of what was actually observed. The channel is, of course, dynamic by its very nature, but this mesh assumes that th e morphologic changes occurring over the past two years have not signifi cantly altered the relationshi p between channel stage and

PAGE 198

184 discharge. Having returned to the field site in October 2002, this is a valid assumption. The contour lines presented in Figure A-1 show clear evidence of thalweg location for the downstream portion of the model domain, but the thalweg location fo r the upper half is not clear. However, the thalweg and water edge locations were precisely surveyed for the upstream portion of this reach during Ju ly of 2001 (Figure A-2). A simple linear interpolation of the scatter points along the thalweg allows for the connection of these points within the model mesh. The remainder of the bed in these areas is fairly flat, and linear interpolation fr om the existing survey points allowed new nodes to be created to improve the nodal density and spacing. The bed characteristics for the downstream half of the model domain, between easting coordinates 744100 and 744300, are more complex. Given that the Scout Camp reach of the Carson River is generally oriented east-west, bed elevation plots were prepared for 50-meter subsections of this east-west distance and for each cross-section. These plots, an example of which is shown in Figure A-3, importantly show the locations of mu ltiple low-flow channels and the high points between these channels. For these plots, the x axis is the northing coordinate (approximately perpendicular to flow direction), the y-axis is the surveyed elevation, and all available survey points are classified and plotted based on which 10-meter group they fall into along the direction of flow (as approximated by easting coordinate). The idea was to be able to see how the bed is changi ng along its longitudinal le ngth, so that this variability could be captu red in the model mesh. The general bed geometry features for the cross-sections are shown in Figure A-4. Cross-section locations are indicated back on Figure 5-1. A few additional points are worthy of note following inspection of the cro ss-sections. Channel bed elevation is lost

PAGE 199

185 4503500 4503550 4503600 4503650 4503700 4503750 4503800 4503850 743800743900744000744100744200744300744400 Easting Coordinate, State Plane 83: Nevada West (meters)Northing Coordinate, State Plane 83: Nevada West (meters) July 2001 March 2001 July Left Water Edge July Right Water Edge July Thalweg Groundwater Wells Figure A-2. Network of survey data points, indicating locatio n of thalweg and water edge Figure A-3. Bed elevation pl ot indicating bed feature ch anges in flow direction 1250.5 1251 1251.5 1252 1252.5 1253 45035604503570450358045035904503600450361045036204503630 Northing Coordinate, State Plane 83: Nevada West (meters)Elevation, NAVD 88 (meters) 744200-210 744210-220 744220-230 744230-240 744240-250

PAGE 200

186 moving downstream, and the channel thalweg be comes more pronounced. All cross-sections show the pres ence of steep channel banks, with cross-section 300 the most telling example of verti cal banks on both sides of the channel. Lastly, bed geometry data is not smooth because of the presence of bed forms such as dunes, depositional sandbars, and multiple low-flow channels. Agai n, the mesh needs to capture the general trends, but within the confines of nume rical convergence. These plots allowed the removal of subjectivity associated with node insertion and deletion. The majority of removed data points were along the five cross-sections. Points retained were deemed as representative of th e cross-sectional geometry, rather than those needed to capture each bed form. Some data points from the July bed survey were also removed near the cross-sections in order to improve the union of the two temporal data sets. The days spent surveying the Scout Camp site provided insight and photographic documentation to incorporate the vertical cha nnel banks and steep slopes into the scatter point network. Additionally, field measuremen ts of bank height were recorded as a component of the erosion study, during which a network of iron stakes were pounded into the vertical portions of the bank. This res ource base allowed for additional nodes to be created along the tops of vertical banks (where data collection was inhibited by the cottonwood trees) and along the ba se of the channel banks (w here the banks themselves often inhibited data collection). Linear in terpolation between two surveyed points was used only as a final resort when other data was lacking. Initial model intention was to include the capability to model flows up to the bank full level, or flows at which channel water depth is above the vertical banks. Many of

PAGE 201

187 these data points along the top of channel the channel banks and on the floodplain were created subjectively or utilizing linear inter polation. Detail was not necessary, and the points simply needed to show a subtle elev ation increase moving away from the channel; the goal here was to identify the flows at which bank full conditions occurred throughout the simulated reach. Given the changing natu re of vegetation (roughness characteristics variable relative to sandy cha nnel bottom), highly variable sl opes, and limited data, these nodes were created using a combination of cr oss-section data where available, linear interpolation, and common sense. Figure A-4. Bed and bank geometry at surveyed Scout Camp cross-sections Improvement of Node Spacing and Drying Properties After triangulating the node s, deleting elements outsi de the boundaries, merging triangles, and adding and dele ting nodes to add channel morphology details, the mesh still needs additional manipulation to add to model stability. At this stage, node spacing becomes of primary importance. While some consideration of this is used during node 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5 1254.0 1254.5 1255.0 0102030405060708090100 Distance from Left Channel Bank, Looking Upstream (meters)Elevation, NAVD 88 (meters) Cross-Section 100 Cross-Section 200 Cross-Section 300 Cross-Section 400 Cross-Section 500

PAGE 202

188 addition, all attempts were made to hold th e locations of surveyed nodes constant. Up until this point, one primary goal was kept in mind: getting the channel geometry contours to be representative of the general observed trends and continuous. The lowest elevation contours follow the channel thalwe g and low-flow channels, higher contours follow the channel bed, still higher contours come together at vertical banks, and the highest contours smoothly traverse onto th e floodplain. Basically the contour lines within the model domain resemble a topographic representation of a large bathtub, with the lowest downstream thalweg el evation representing the drain. Recall that quadrilateral elements are pref erred for two reasons: they make a more concise mesh for faster solutions, and quadrilat eral elements are numerically more stable, particularly when dealing with areal wetti ng and drying (SMS Tutorial: Version 8.0, Lesson 3.4). Lines of quadrilat eral elements should run para llel to the boundary so that the mesh will wet and dry parallel to the boundary (SMS Tutorial: Version 8.0, Lesson 3.6). Referring back to Figure A-1, several si gnificant obstacles to developing elements of this pattern are noted. Th e algorithm used by SMS to triangulate nodes into elements always connects the three nodes closest to one anot her into a triangular element. The algorithm to merge triangles into quadrilateral s only merges triangles when all resulting quadrilateral angles exceed a user -input value, th e default being 65 Quadrilaterals are then created in the or der of increasing minimum angle. Simply put, given the variable node spacing in this mesh, the quadrilateral elements created will not run parallel to the mesh boundary and will include several triangular elements internal to the mesh. The decision was made that the mesh would be completely triangulated, including only triangular elemen ts, and the quadrilatera l elements would be

PAGE 203

189 created one at a time using the features of the FESWMS user interface. The modified triangular mesh, including all node additions and deletions un til this point, is shown in Figure A-5. Also shown are th e underlying contours for this triangulated mesh. Note the continuous nature of the contour lines and c onformity to field observations. Importantly, the mesh will be able to transport fluid through the model domain, at least in theory. elevation 1250.17 1250.67 1251.17 1251.67 1252.17 1252.67 1253.17 1253.67 1254.17 1254.67 Figure A-5. Modified tr iangular mesh after node addition and deletion The user interface of the FESWMS module o ffers several features useful to the development of a quadrilateral mesh and the improvement of node spacing. Useful at this time is to visualize the desired end-product. The rectangular mesh will need to dry parallel to the boundaries, a nd to enhance numerical stab ility, the use of triangular elements within the mesh interior will be minimized. The approach was to visualize a network of channel transects, perpendicular to the direction of channel flow. By dividing the model domain in this fashion, the overall mesh would be built one transect at a time. Flow direction is approximated by the cha nnel thalweg. As such, getting the thalweg right was the starting point of this process. With a fair ly extensive network of nodes

PAGE 204

190 already in place and no large spatial gaps, linear interpol ation was used to add nodes where needed so that they would fall along th ese transects. With the nodes in place, the split / merge tool is FESWMS was used to merge each set of two triangles into the desired quadrilaterals. The progr ession of this procedure is shown in Figure A-6, Figure A-7, and Figure A-8. Initially, the process was applied only to the channel bottom, working from the thalweg outward and dow nstream to upstream. Mesh points were deleted and inserted as needed, but data poi nts representing actual survey observations were retained whenever possible. Admittedly, the process is tedious and time consuming, but it is made simpler with the underlying cont ours in place. Inse rted nodes, in other words, conform to the existing lay of the land. Triangular elements are needed along bends in the channel to maintain uniform node spacing. Triangular elements are also included in a few locations to preserve details of the channel geometry. After the main channel was completed, the process was repeated with the floodplain. Here, the lack of data required the insertion of many more nodes. The linear interpolation was used to insert multiple nodes between two known points, with the exact number based on extension of transects to the model boundaries. The process is completed working outward from the cha nnel banks to the model boundary. The end result, the completed quadrilateral mesh, is shown in Figure A-9. For reference, the original survey points have al so been shown. Some nodes were moved slightly, using the unlock node option, to improve uniformity of node spacing. These minimal alterations in node locations will not alter the final solution to within the e xpected error associated with the data itself. Wherever possible, original survey node locations were retained. The

PAGE 205

191 framework is now developed for a quadrilateral mesh that will wet and dry parallel to the boundaries. elevation 1250.17 1250.67 1251.17 1251.67 1252.17 1252.67 1253.17 1253.67 1254.17 1254.67 Figure A-6. Downstream main chan nel, quadrilateral mesh creation elevation 1250.17 1250.67 1251.17 1251.67 1252.17 1252.67 1253.17 1253.67 1254.17 1254.67 Figure A-7. Middle main channel, quadrilateral mesh creation

PAGE 206

192 elevation 1250.17 1250.67 1251.17 1251.67 1252.17 1252.67 1253.17 1253.67 1254.17 1254.67 Figure A-8. Main channel comple te, quadrilateral mesh creation elevation 1250.00 1250.70 1251.40 1252.10 1252.80 1253.50 1254.20 1254.90 Figure A-9. Completed qua drilateral mesh with bathymetric contours

PAGE 207

193Mesh Corrections for Hydrodynamic Simulations With the critical geometric features of the channel, including the thalweg, bed high points, steep vertical banks, rip-rap, and the floodplain, consiste nt between model and reality, the next process is getting the model to work and convey water. Water conveyance is accomplished through the use of boundary conditions. The finite element mesh was developed so that the upstream and downstream boundari es each approximate transects, perpendicular to fl ow direction. The create node string feature of the FESWMS toolbox is used to create a nodestring connecting all nodes along these two open boundaries, boundaries at which flow enters or leaves the mesh. Four options are available within FESWMS for boundary definiti on: specified flow / WSE, supercritical outflow, computed WSE, and St orm / Tide WSE (note that WSE is an abbreviation for water surface elevation). These boundary condi tions will be applied at the upstream and downstream cross-sections of the Carson River model domain. The remaining model boundaries are, in essence, vertical walls These boundaries may be applied roughness values and serve as no-flow boundaries. A required condition for the starting a FESW MS simulation is the submergence of all nodes within the model domain. The initial water surface elevation must, therefore, be greater than the elev ation of the highest node within the mesh. For the Carson River model, flow within the model domai n will be subcritical (Fr = V / (gD)1/2 < 1.0). The avoidance of supercritical flow conditions at the specified upstream and downstream boundaries is the principal determinant as to the minimum initial water surface elevation that can be specified. Given that flow through the model domain will be primarily subcritical in nature (except in cases where flow depths approach zero as elements are very close to dry or recently rewet) and that tidal cond itions (as needed for a river

PAGE 208

194 entering an estuary or ocean modeling) are not being simulated, boundary conditions will be of two possible types: head boundary c ondition or flow boundary condition. FESWMS allows flow boundary conditions to be specifi ed as constant or dynamic. While head boundary conditions are also spec ified as constant or dynamic, head values at a boundary nodestring can be manually input, read from a rating curve (with the independent variable flow specified), or specified using a friction slope. The fric tion slope option allows for a non-level water surface along the boundary. The us e of the rating curve or friction slope options, do introduce additional problems as will be later discussed. Development of a Simple Linear Model In order to gain experience with SMS, determine the best way to parameterize the model, and develop insight for further mesh refinement, a smaller and simpler linear model was prepared. The model captured the in ner thalweg, small sloping bed, vertical banks, and floodplain which characterize th e Carson River and the prepared mesh. However, the number of elements for the simple model was only a fraction of the complete model, allowing for reduced com putational time and the opportunity to focus initially on the big picture. Clearly, the most obvious problem area was the steep (almost vertical) banks. A profile view of this simple model, including the same quadrilateral elements as the complete detailed model, is found in Figure A-10. Cross-sectional shape and dimensions were taken from Figure 3-1. The simple model was constructed following the same procedure as the detailed mo del, with the notable exception being that the input scatter points were created with re spect to an arbitrary datum and based on the dimensions indicated in Figur e A-10. A channel bottom slope of 0.001 (1 meter per 1000 meters of channel length) was used along the 500-meter straight channel reach. Bottom slopes and cross-sectional geometry remain uniform throughout the model domain.

PAGE 209

195 Figure A-10. Uniform Cross-Section Dime nsions for Simple SMS Model Mesh All node points were not directly calculat ed for the input data points. Rather, several useful features with FESWMS were used to add additional elements. The breakline feature allows for the creation of new elements by transforming mid-side nodes (selected as part of an input nodestring) into element corner nodes. The refine command allows for the division of select ed quadrilateral elements into four smaller elements. Both options apply linear interpolation to existing node positions and elevations. SMS Model Insights Provided by Simple Linear Model Repeated experimentation using this simple model resulted in several important and useful discoveries about the modeli ng system operation and limitations. This experimentation yielded information a bout specifying boundary conditions; and the effective and efficient hand ling of steep channel banks. Specification of boundary conditions First, it is necessary to review the specification of boundary conditions. For a simple model domain, with uniform bottom sl ope and uniform cross-sectional geometry, 1.50 22.50 22.80 0.20 m 2.30 m 4.00 m 4.30 m

PAGE 210

196 the model can easily simulate the follow ing boundary condition specifications: upstream flow with downstream head, or upstream head with downstream head. This introduces a potential problem with reference to the m odeled reach of the Carson River. While the USGS gaging station at Fort Churchill allows for a reasonable approximation of the flow at the model upstream boundary, no data is available for the downstream boundary. No stage recorder was installed at the field si te, and no stage-discharge relationship is known. While the user may specify an upstr eam flow boundary with a downstream flow boundary, two details should be realized: travel time is a function of flow magnitude and velocity, and a numerical model specified in th is fashion will be ex tremely unstable. The interface does allow for the sp ecification of a ra ting curve boundary condition, but model reference to this as a computed water surf ace elevation can be misl eading. In fact, the model will not independently compute this downstream head elevation based on an upstream flow condition. Rather, the rating cu rve relationship (stage as a function of discharge) must be specified and the consta nt or transient discharge values must be provided. In other words, downstream head values must be known either through direct field measurement of stage (or depth) or di rect field measurement of flow. The friction slope option can only be applied across th e downstream boundary, ra ther than along the downstream boundary. This option allows for a non-level water surf ace to be specified along the boundary nodestring, and the input value does not repres ent the water surface slope passing over the boundary out of the model domain. The lack of stage data available for the modeled reach of the Carson River is therefore a pr oblem in need of a solution.

PAGE 211

197 Recall that the detailed mode l for the Scout Camp reach of the Carson River is approximately 500 meters in length. Th e downstream model domain boundary could presumably be extended far enough downstream so that the effects of the downstream boundary would not be felt within the curren t model domain. This would essentially involve creating new scatter poi nts to extend the m odel mesh into a re servoir, within which the head could remain constant. Of cour se, given that no survey data is available for further downstream and the uncertainty as to how far would be far enough, this option was not selected. Rather, the model has a usef ul way to sidestep this problem. Model solutions predict a water surface elevation, wate r depth, and velocity vector for each node in the model domain. Using the solution veloc ity vectors and depths the model calculates flow across any nodestring. Such nodestri ngs can include the upstream boundary, downstream boundary, or say, a nodestring across the channel at the well cross-section. With no stage data available for th e downstream boundary and no means to independently calculate and veri fy this relationship, it was determined that the use of both upstream and downstream head boundary conditions were the best means to simulate the hydraulics of the channel. The model output co uld then be used to develop rating curves for the modeled reach, specifically a rati ng curve to predict the stage-discharge relationship for the vertical bank adjacent to the groundwater wells. The repercussions of this decision will become more evident as this discussion progresses. Of particular interest is the choice of water surface sl ope and the linkage between the respective boundary head conditions. Model handling of steep vertical channel banks The steep banks of the simple model also posed problems. A slope is necessarily applied to these banks. Within the model mesh, two points ca nnot be located at the same

PAGE 212

198 latitude and longitude in space. Node elevations for the base and top of these banks were often not each available at a given location due to aforementioned surveying difficulties with point acquisition. Rather, some of th ese points are based on collaboration between field measurements, photographs, and recollec tion of the modeler. The SMS model will allow creation of these steep elements, but wetting and drying of these elements introduces complications. A discussion is necessary with regard to how element wetting and drying is performed within the SMS model. The finite -difference solution scheme predicts water surface elevations for all nodes within the model domain. FESWMS parameters set in the FESWMS model control determin es how elements are checked for their wet or dry status. Two specific parameters need to be defined. Default storativity is defined as the water depth below ground elevation at which a node is considered to be dry. When a node becomes dry, all elements attached to the dry node are considered dry. A zero value for storativity indicates that a node will become dry when the wa ter surface elevation is equal to the nodal ground elevation. A positive value indicates that the element will only dry when the computed water surface is some height below the node ground elevation. Values in the range of 0.0 to 0.1 meters we re used throughout the m odeling process, with higher values best suited for elements possessing the steepest slopes. If the numerical model determines that an element has dried out, any flow previous ly passing through that element is passed over to elements that rema in wet. The second notable parameter is the depth tolerance for drying. The element drying / wetting toggle is used to control how FESWMS handles changing water surface eleva tions. The toggle needs to be checked in the FESWMS parameters dialog box if elements are to be turned either on or off during a

PAGE 213

199 given simulation. If turned on, the input depth tolerance is used to determine if a partially submerged element should or should not be in cluded in the comput ational analysis. In this case, the water surface elevation for an adjacent element is higher than the minimum node elevation for the bordering dry element, presumably those nodes along this wet / dry boundary. When the water depth over any node of a dry element exceeds the depth tolerance threshold value, a previously dry elem ent will be considered wet, at least for the purposes of the given iteration. If too many el ements get classified as dry in a single iteration, the water surface elevations in the remaining elements may go up enough to incorrectly make the dry elements wet agai n. The shifting back and forth between wet and dry elements results in numerical instab ility in the solution. Such instability is particularly evident for steep elements, as shown in Figure A-11. As an illustrative example, consider node 1 going dry when the water surface drops below the node elevation by the storativity depth for one ite ration, but the water depth in the adjacent element (sharing node 3) exceeding the dept h tolerance for drying during the next iteration; the hatched porti on will oscillate between we t and dry and the underlying element may be unable to converge or ev en diverge, depending on other adjacent elements. The repeated shifts of elements from we t to dry will prevent model convergence. In cases where several adjacent elements si multaneously undergo this shifting, the magnitude of the oscillations will often progr essively increase with each iteration. The resulting changes in nodal velocities and wa ter surface elevations will eventually cause model divergence and a model crash. The pr oblem is enhanced further during dynamic simulations in which the water surface is already changing from time step to time step.

PAGE 214

200 Figure A-11. Element Wet / Dry Oscillation The problem here is the maximum iteration cr iteria. If the model has neither converged nor diverged during the time step, the model will advance to the next time step. Certain elements which should be wet will be dry, and certain elements which should be dry will be wet. Typically, it is only a matter of a fe w iterations before model divergence occurs. Evidence of this numerical instability is found from steady state simulations of the simplified SMS mesh. Because flow is able to be lowered from floodplain levels completely down to zero, attempts were made to develop rating curves from a series of steady state simulations. The head boundary c onditions were incrementally lowered for a constant water surface slope, with equilibri um flow conditions recorded. The simple mesh contained only a single element along th e steep banks, but the drying of these elements causes an inconsistency in the rating curve. For a typical rating curve, increasing depth of flow is associated with increasing flow. However, the sudden drying 9-Noded Quadrilateral Element Nodes Side View of Element 1 2 3 1 2 3 Oscillating Between Wet and Dry

PAGE 215

201 of theses elements when the top nodes are dry disallows flow through these elements. Equilibrium flow increases for a decreased de pth of flow. This is illustrative of how sudden drying of a large portion of the elements in a model domain can significantly alter flow conditions. For the Carson River m odel, the bank elements would not dry simultaneously as they do here and the bank elements transport far less flow than the remaining mesh elements, but the probabi lity of model divergence for a dynamic simulation in which several bank elements become dry is suggested. FESWMS Model Parameterization: Element Wetting and Drying Before running a simulation using FESWMS a conceptual unde rstanding of how the model performs computations combined with a few observations concerning the raw survey data allows for prediction of likely causes of model failure and modify the mesh in advance of their occurrence. Recall the model requires all mesh nodes to be initially submerged. In the model mesh shown in Figur e A-9, the water surface elevation at both boundaries was to be lowered at a constant rate. Head boundary conditions would be used, and the water surface slope would be held constant as the water surface is incrementally lowered in a step fashion. Two specific probable divergence problems were identified. First, the survey ed elevation data includes hi gh and low points along the lower floodplain. The morphology surr ounding the tops of the chan nel banks is not at all smooth, including hills and valleys on both a large and small scale. Elevation inconsistencies will ensure a jagged wetti ng and drying boundary. With the model only allowing flow to pass through “wet” elemen ts, the flow pattern from upstream to downstream will be effectively interrupted fo r these regions. As the model attempts to compensate through changing velocities and water depths, divergence is inevitable. Secondly, the model will experience divergence due to the aforementioned oscillation of

PAGE 216

202 elements from wet to dry. The drying of a single steep bank element from a single node becoming dry would often result in an adj acent water surface computation higher than the dry node itself, and certainly an ad jacent water surface depth sufficient to cause re-wetting for any small value of the depth tolerance parameter. To circumvent these problems and retain the ability to simulate flows above the bank-full river stage, nodes were inserted to sub-divide single steep bank elements (two corner nodes at both top and bottom of a give n bank) into between three and five smaller elements with vertical dimensions on the order of one-half of one meter. The depth tolerance for element re-wetting is also set to be greater than height of these steep bank elements. Of course, how realistic is th is? Given that small magnitude numerical fluctuations are likely during the course of any solution, th e modeler is effectively not allowing the model to self-correct these probl ems and re-wet an element that should be wet. In addition to assisting the model to run, justification is found in the fact that the flow through a few bank elements mistakenly turned off is negligible by comparison to the flow through the mesh as a whole. This is definitely valid at the high flows required for extensive bank contact, and the resulting water surface elevations will not be altered by any amount greater than the errors asso ciated with the surv ey data points, node elevation modifications, or we ll data precision. In summation, this simplification should allow for simulation of a falling river stag e. The jagged interface between wet and dry elements and the resulting numerical divergence proves to be a problem difficult to solve. This is a simply by-product of the nature of the channel morphology, within which elevations gradients do not follow smooth rising or falling curves. This will be a problem in need of a solution, or at least some means of control.

PAGE 217

203Modeling of Steep Channel Banks General steps prior to start of hydrodynamic simulations With the first round of mesh refinements in place, the mesh should be capable of simulating channel hydrodynamics. Before running the FLO2DH model used by FESWMS, a few additional steps do need to be completed. First, the model boundary should be smoothed. To prevent loss of ma ss through the jagged boundary of a quadratic mesh, smoothing will move the mid-side nodes along the model boundary while preserving the locations of corner nodes (SMS Tutorial: Version 8.0, Lesson 3.7). Smoothing requires the creation of a nodestring in a counter-clockwise direction between the two extreme nodes along the upstream boundary condition nodestring. The select nodestring tool is used to se lect this nodestring, followed by the use of the ‘Elements / Smooth Nodestring’ command. The process of creating and editing a finite-element mesh causes the node and element numbering to b ecome almost random, thereby increasing the size of the matrices required by the finite element analysis codes. Renumbering the mesh restores order. After selecting the node string at the upstream boundary condition, the band width option is selected from the ‘Ele ments / Renumber’ option, and the mesh is renumbered. In all instances involving changes of nodes along the mesh boundary, the mesh boundary should be smoothed. Similarly, the addition or deletion of nodes from the mesh alters the node numbering, and renumberi ng is always the last step of the mesh creation process. FESWMS also includes a mesh quality feature. This feature visually identifies potential problems with the finite-element mesh layout. When this option is checked under ‘Display / Display Options,’ an elemen t is highlighted in a color corresponding to the criterion which it violates. Mesh quality affects the validity of a finite element

PAGE 218

204 analysis. While not all problems are id entified, problems known to cause model instability are identified and can be easily corrected using this quality check. Available quality checks include the following: minimum interior angle, ambiguous gradient, concave quadrilaterals, maximum slope, and element area change. A brief discussion of each mesh quality check is herein provided (SMS Help Menu). Minimum interior angle warns of skinny triangles and skinny diamond-shaped quadrilaterals. Ambiguous gradients occur when two diagonal corner nodes of a quadrilateral are higher than the other two corner nodes, forming a saddle point along the middle of the element; these elements are undesirable because a drop of water at the middle of the element could flow downhill in two different directions. Concave quadrilaterals possess large interi or angles and should never be included in the mesh. In the flow direction, a steep element slope i ndicates the possibility of supercritical flow; although FESWMS supports supercritical fl ow, supercritical c onditions along boundaries specified as subcritical may cause the model to crash. In the direc tion perpendicular to flow, a steep element indicates possible wet / dr y shocking. It is sugge sted that the area difference between two adjacent el ements not exceed 50 percent. Before any model simulation run, mesh quality is checked and problems are remedied. Minor dragging of nodes to a new location (limited by elements formed by bordering nodes) can be useful to alle viate minimum interior angles, concave quadrilaterals, and element area change wa rnings. Other remedies include addition of new nodes by linear interpolation to create sm aller elements, node deletion to re-form elements, switching element shape from triangul ar to quadrilateral (or the reverse), or switching diagonals for two triangular elemen ts within a quadrilateral. Maximum slope

PAGE 219

205 warnings perpendicular to flow are co mmon within the Carson River mesh and unavoidable given the nature of the channel morphology. Throughout the mesh modification process, the crea tion, deletion, or movement of nodes is always followed by a mesh quality check. It is a good recommendati on to perform a quality check before any simulation, as it will save the loss of time associated with a foreseeable and hence avoidable model crash. Identification of mesh problems after a hydrodynamic simulation One clear advantage of the simple linear model was that all node elevations were effectively inserted by hand. The channel cross-section shape does not vary moving downstream. Even though the bank elements we re steep (possessing ne arly a 6:1 slope), the mesh elements were well constructed. Un fortunately, the steep bank elements within the Carson River model do not exhibit this same uniformity. In most cases, surveyed elevations along the tops and bottoms of banks were linearly interpolated to generate additional points along these characteristic feat ures. Linear interpolat ion was additionally used to subdivide the bank elements into sma ller elements in an attempt to prevent model divergence caused by simultaneous drying of elements, thereby preventing flow through the element. Given that the surveyed eleva tions do not smoothly rise or fall along the banks, element areas often differ substantiall y between adjacent element, and rows of bank elements do not typically follow the underlying contour lines. Nodal elevation changes, slope changes from one bank element to the next, and the sudden increase in flow depth during the mesh transition from floodplain elements to bank elements cause significant instabilities within the finite diffe rence solution algorithm. Elements subject to these instabilities can be easily identified in two ways. One way is to run the FESWMS FLO2DH model, allow the model to crash, import the solution final (the .flo file) and

PAGE 220

206 view the mesh properties for the time step s immediately precedi ng the model crash. Nodes within unstable elements will exhibit trends during these pre-crash time steps including the following: water surface elevations (or water depths) well above or below the range imposed by the boundary conditions, velocity magnitudes significantly higher than other elements within the model domain, or velocity vectors f acing in directions not consistent with the principal direction of fl ow. These problem areas are easily identified in the solution window, as the contour range is set by default to automatically adjust based on the range for all nodes in the mesh solu tion. When the model is set to color fill, versus only contour lines, over the spectrum of colors, distinct color differences will jump out for these diverging nodes. Vectors fo r velocity solutions can also be shown for all nodes, making inconsistent velocity values literally jump off the page. Examples of diverging solutions are show n in Figures A-12, A-13, a nd A-14. These three figures actually represent the same solution. Note the contours as indicated in the legend. Recalling that the highest nodes of the mesh have elevations of 1254.45 meters, it is noted that all nodes in the model mesh are submerged under greater than 20 meters of water. Considering the channel morphology, as indicated in Figure A9, water depths are predicted to be close to uniform within th e main channel, excluding the increased depth at the downstream thalweg. Figure A-12 cl early shows a diverging solution with physically-unreasonable water depths, most notably along areas characterized by steep channel banks. The diverging velocity magnit udes and vectors for this same solution are evident in Figures A-13 and A-14. In the case of Figure A-14, the finite-element solution is predicting velocity vortices around the areas of extreme water surf ace (related to depth

PAGE 221

207 and bed elevation) variation. These are physic ally unrealistic solutions that are output by the model, indicating the importance of quality control checks by the user. The second means of identif ying potential diverging node s is through examination of the output file (the .prt f ile). Poorly formed elements may result in a negative or zero determinants that will lead to unacceptabl e solutions. These elements are identified during each model iteration by the following warning message: Determinant <= 0 in element X at integration point X. Divergi ng nodes may not immediately cause the model to crash. The maximum number of iterations for all time steps is 99, and the number of allowable iterations can be specified as a ny number less than 99. A few diverging nodes may not cause the model to crash, but the model will be unable to converge. With increasing numbers of iterations, however, the divergence of one node will influence adjacent nodes, and eventually the model wi ll crash. The best solution around this problem is to inspect each final model solu tion to check for inconsistencies in nodal Figure A-12. Physically-unrealistic wa ter depths along steep channel banks 1278 1277.8 water depth 15.0 17.0 19.0 21.0 23.0 25.0 27.0 29.0 Model divergence at steep channel banks

PAGE 222

208 1278 1277.8 velocity mag 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 Figure A-13. Model divergen ce, velocity magnitude 1278 1277.8 velocity mag 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 Figure A-14. Model diverg ence, velocity vectors Model divergence at steep channel banks Model divergence at steep channel banks

PAGE 223

209 elevations or velocities. In addition to substantially increas ing model run time, final water surface elevations in the solution may not be physically reasonable. Where these diverging nodes are located, th e elements need to be recons tructed, either through minor node relocation to improve uniformity of adjacent element shapes or through adjustment of node elevations to smooth the elevation gradients. Attempts to Fix the Problem Areas Iden tified by a Hydrodynamic Simulation Mesh corrections are an iterative proce dure. All model nodes must be submerged for the initial conditions. Initially, head bounda ry conditions were used for simple steady state simulations. When the maximum number of iterations was reached or the model crashed, the output .prt files were examined a nd diverging nodes record ed. The .prt file is created for all model runs, re gardless of whether a model crash occurs. The DOS window in which FLO2DH runs only tells the model user the number of errors and warnings, explaining nothing about model divergence or a model crash. By zooming in on these diverging areas, clicking on node s to view elevation propertie s, and utilizing 3D viewing angles with exaggeration in th e height dimension, the dive rging nodes were corrected one at a time by dragging nodes or inserting node s to create uniform node spacing, minor modification to node elevations to smooth gr adients between adjacent nodes, or some combination of both. The model would then be run again subject to the same boundary conditions to see if the correction had work ed properly. The mesh would either be corrected again or the boundary conditions wo uld be lowered and the procedure repeated. Most of these diverging nodes we re present along the steep por tions of the vertical banks, but steep portions of the inner channel also posed some divergence problems that required correction. Many of the divergence problems were simply a by-product of using raw survey elevations, resulting in abrupt elevation changes or poor node spacing. These

PAGE 224

210 trouble spots were able to be corrected so that the steady state simulations would not crash and flows through the model domain a pproached constant values. However, this only held true for flow depths in excess of ten meters above the highest node elevations in the mesh. As the water surface was gra dually lowered using the boundary conditions, regardless of the rate of lo wering, supercritical flow warnings would arise for nodes along the steep portions of the mesh. Within a few iterations of thes e supercritical flow warnings reaching nodes along the boundary conditions nodestrings, the SMS model would crash. The problem of steep banks in a finite element model One common thread tied these problem nodes together. These nodes were all found in steep portions of the mesh, either along ch annel banks or channel thalweg. Other nodes did cause diverging solutions at first, but these nodes were ab le to be corrected using the above-discussed procedures. Given that the goal is to lower the water surface elevation to a level representing close to zero flow in th e channel, the inability to lower the water surface elevation to even the levels of the highest node without model failure posed a definite problem. Boss International hosts an online fo rum through which SMS model users can direct questions to technical support sta ff or post advice or co mments from personal experience with the model. In posing a question to post to the forum, the following important information was included: general mesh characteristics, including the nature of elements along the steep channel banks; bounda ry conditions used, and a description of warnings preceding model failure. Feedback was collected from a combination of suggestions from other SMS model users, prev ious postings on the site, and a review of the model mesh by members of the Boss Inte rnational technical support staff. It was

PAGE 225

211 verified that steep elements cause converg ence problems and noted that these elements generally do not carry much conveyance, th ereby providing justific ation to ignore these bank elements when necessary. FESWMS assumes that mesh boundaries are no flow boundaries, able to be defined as either frictionless or not. Additional suggestions included lowering of the relaxation fact or, specified under ‘FESWMS Control / Iterations,’ to a value of 0.01. The range for this value is 0 to 2, with a default value of 1.0. According to documentation within th e SMS help menu, higher values should be used if a model demonstrates slow convergen ce. Smaller values can be used to increase numerical stability if a model diverges. A lowe r value will give a sl ower but more stable solution run. The technical s upport team did note some mesh problem areas, such as poor element construction along one channel bank (Figure A-15), but fixing of these nodes and elements still did not result in model crash pr evention. Other potential alternatives were discovered by searching through the forum arch ives. Potential alternatives included the following: reduced nodestring length along th e upstream and downstream boundaries to prevent boundary condition elements from drying, straightening of the boundary condition nodestrings to ensure perpendicularity to flow di rection, shortening the overall mesh length to check if convergence is attain ed, turning off the wet / dry toggle for these submerged conditions, defining different roughness values for floodplain and main channel flows, and utilizing an upstream flow boundary condition. Each approach failed to solve the problems arising from the steep channel banks. The biggest piece of advice taken from the experience of other modele rs was the following: if modeling of the floodplain in not critical, then eliminate the overbank areas and steep elements along the

PAGE 226

212 channel edges and use a semi-slip or no-s lip boundary with a ve rtical wall roughness appropriate to the bank condition. 1278 1277.8 Figure A-15. Poor element construction along channel bank, 3D model illustration A new approach for including of ve rtical banks in the model mesh After repeated experimentation and consul tation, it was evident that an alternate approach would be needed to solve the m ounting problems related to the steep channel banks characteristic of the Carson River ch annel. Steep elements bring about sudden changes in water depths and velocities betw een adjacent nodes within steep elements, and the finite-difference algorithm of FESWMS inevitably moves toward divergence and a model crash. Realizing that the simulation of floodplain fl ows could result in assumptions that would bring the model results into question and noting that the maximum Carson River streamflow for the period covered by th is analysis was below the bank-full level (7555 cfs at Fort Churchill, estimated usi ng RIVMOD), it was decided that the model mesh would be significantly modified to effectively remove the floodplain elements. Besides, most of the floodplain nodes were not field surveyed and were only later added. Problem with mesh elements not being uniformly distributed in elevation

PAGE 227

213 Additionally, FESWMS offers one feature that would prove pa rticularly useful for this change. The model mesh includes a finite spatial area defined with a boundary. This boundary is a vertical wall with no upper lim it. Flows magnitudes causing contact with the boundary are therefore not subject to elemental drying, except when water surface elevations drop below the storativ ity depth specified at the base of this imaginary vertical wall. A vertical wall shear coefficient is specified within the FESWMS parameters dialog box. Based on the availability of no data, it is assumed that the shear stress applied to the bottom of the channel will be the same as the shear stress applied to the vertical walls. Semi-slip conditions are selected in the FE SWMS control box and a value of 1.0 (100 % of bottom stress) is entered with in the FESWMS parameters box. Mesh modification required careful pla nning and frequent saving of work. The model interface does not include an undo butt on. The deletion of nodes and elements causes the model boundary to move inward to ward the main channel. Nodes cannot be inserted outside of the model boundary without starting over at the triangulation phase of mesh development, at which the outer boundaries are determined. Accidental node deletion, resulting in an undesired boundary, ca n therefore pose a problem. Nodes within the boundary can be created and relocated, a nd node elevations can be recovered from original survey data, but typically it is easie r to revert to a saved version of the mesh, from prior to any undesired mesh modificati on. Due to the small size of many elements, notably along the steep portions of channe l bank, nodes were removed one at a time. Prior to node and element deletion, pla nning was needed to identify these nodes and elements. The established goal was to provide a mesh able to accurately simulate flow conditions up to bank-full conditions. First, the locations of all steep, nearly vertical

PAGE 228

214 banks were identified within the detailed mesh Photographs and field data were reviewed as an additional accuracy check. Unfortunate ly, vertical banks do not characterize the entire 500-meter surveyed reach of the Carson River. Portions of the Scout Camp reach included a gradual slope from channel bed to floodplain or the presence of rip-rap for erosion control. Regions characterized by ve rtical banks are identified with ovals in Figure A-16. These vertical banks would b ecome vertical wall boundaries for the final mesh. The disconnectedness of these vertical features meant that all mesh boundaries could not be moved uniformly closer to the ch annel. Rather, the top of the vertical banks was used as a reference. The extent of mesh retained between segments of vertical banks was based on one criterion: making certain that the highest elevations in these less steep areas remained higher than the tops of the vertical banks, as k nown from the present model mesh. This consideration ensures that the model will have the capability to accurately simulate river flows up to th e bank-full threshold. Beyond this singular criterion, mesh refinement was based on both providing a smooth model boundary for improved wetting and drying performance and also minimizing the number of elements, so as to decrease computation time. The first cut of mesh refinements to create true vertical banks is shown in Figure A-17. All mesh boundaries represent vertical walls with roughness values equivalent to those of the channel bed. Roughness values were assumed equivalent to bed roughness values used in RIVMOD for the IFMTFM, a value of 0.045. Channel banks not characterized by steep banks are maintained with transitional areas between the two bank types. Wh ere vertical walls are presen t along the channel, the node elevations along these vertical walls are the su rveyed or estimated elevations at the base of these features.

PAGE 229

215 elevation 1250.00 1250.70 1251.40 1252.10 1252.80 1253.50 1254.20 1254.90 Figure A-16. Areas along Scout Camp reach characterized by steep vertical banks elevation 1250.00 1250.67 1251.33 1252.00 1252.67 1253.33 1254.00 1254.67 Figure A-17. Mesh alteration to incorpor ate vertical walls along mesh boundaries

PAGE 230

216Steady-State FESWMS Simula tions, Elements Submerged Setting of Model Parameters With the mesh framework now in place, the next phase of mesh development revolves around lowering the water surface to simulate drying conditions. The ability to construct a model mesh in which elements dry smoothly would prove a difficult task. A couple of questions first need to be answered Is element re-wetting important? At this point, the intention remained to have the FESWMS model simulate Carson River flows over the entire period for which water table data from the wells was available. With channel flows during this period both risi ng and falling, elemen t wetting and drying would both need to be considered. A reasonabl e range for storativity and depth tolerance for drying was determined to be 0 – 10 cm and 10 – 30+ cm, respectively. In simplistic terms, if the storativity is 0.0 and the depth tolerance for drying is 5 cm, the node goes dry (causing all elements that use the node to go dry) when the computed depth reaches zero. The node and associated elements remain dry until the computed depth gets above the depth toleran ce value of 5 cm. If the storativity is a positive value larger than zero, the node does not go dry until the computed depth drops below the node elevation by the storativity depth. Less flow, however, is transmitted through the elements attached to the node if the depth is be tween zero and the storativity value, an effective negative water depth. Storativity depth must be carefully parameterized. If the computed water su rface drops below the ground elevation of all nodes in an element without falling below any one node by a value larger than the storativity depth, then a “wat er surface elevation < 0 for node X in element Y” error will result, causing the model to crash. This f unctionality suggests that upper end of the storativity range will be most useful to simula te drying of elements with moderate slopes,

PAGE 231

217 where the elevation shift from channel bed to floodplain levels is no t as abrupt as where vertical banks are situated. Storativity values of zero are essential when stage levels move below the banks and onto the flatter channel bed. The trade-off here is the premature drying (and inability to transmit flow) of so me elements within the channel thalweg, where slopes are comparable to the non-vertical channel banks. Similar logic is used in the setting of th e depth tolerance for drying. If the water depth along any side node of a “dry” elemen t exceeds the depth tolerance value, the element will re-wet during the next iteration. Of course, th e storativity value will be checked during the next iteration, and th e element will likely again become dry. Additionally, dry elements adjacent to the newl y wetted element can also be rewetted if the water depths are sufficient. This dryi ng and rewetting can result in wet elements disjoint from the wet portion of the mesh. Given that the model has no means to remove water in disconnected “ponds” from the mesh, a sufficient volume of trapped water can cause several nodes to oscillate between wet and dry, often causing water surface elevations and velocities to move outside of a reasonably expected range. As the model is merely performing a series of computations without any logic checks in place, the model may not crash. The model will, however, not converge, and the errors will propagate from one iteration or time step to the next. It is critical, therefore, to visually inspect all final solutions for reasonable values and no disjoint wet elements. Under conditions where wet elements do not become disjoint, but only oscillate between wet and dry, the problem is the increase in computational time caused by the failure to converge. As the model attempts to compensate and force c onvergence, divergence can result. Through a trial and error procedure, depth tolerances gr eater than 30 cm were found to work for the

PAGE 232

218 stages in which moderately-sloped banks elements remain wet, while lower values were found only to work for low flow levels close to the channel thalweg. With the storativity and depth tolerance lim itations in mind, it was decided to refine the model mesh under conditions of drying, whereby the head boundary conditions are always dropping. At this time, one was not in terested in allowing the model to come to steady state through use of step function head decreases. Simple linear decreases were applied to the upstream and downstream h ead boundary conditions. Depending on which elements were subjected to drying, variable rates of drop were n ecessary. The balancing act here was to lower the stage as quickl y as possible, but wi thout a loss of model stability. Stages over the steeper portions of th e mesh were able to be lowered at rates of 0.1 meters over 3 or 4 hours, as relatively few elements dry for a given time step of 15 minutes. As flows dropped to levels in which channel bed elements become dry, slower rates were needed. With the approach in place, all that remained was the determination of the head gradient between the boundary c ondition nodestrings. For mesh refinement purposes only, as opposed to generation of fi nal results, a gradient of 0.2 meters was chosen. The belief was that a small gradient (0.2 meters of head loss over approximately 500 meters along the thalweg) would result in a smooth drying pattern of elements around mesh bends and steeper mesh elemen ts. A node causing model divergence or a model crash under the most passive of conditions would clearly result in model problems under higher velocity conditions, and therefore represented a portion of the mesh in need of adjustment. Convergence Problems Present under Submerged Conditions While looking to lower the water surface through the model domain from completely submerged conditions to levels at which steep banks are no longer contacted,

PAGE 233

219 goal number one was the removal of all warn ing messages from the model output file (the .prt file). The earliest model simula tions used boundary condition head levels of 1265.0 and 1264.8 meters, respectively, at the upstream and downstream boundary nodestrings. The highest node el evation within the model me sh is only 1254.45 meters. All mesh nodes are clearly submerged, as ma ndated for a cold start (no previous solution at all nodes) simulation. A simple steady state simulation indicated 24 nodes which produced “determinant < 0” warning message s. The FESWMS manual indicates this as caused by poorly constructed elem ents, but the output file does not indicate any course of action. The convergence problems associated with these nodes are evident in the steady state solution. Node velocities and water depths will appear to be inconsistent with those of most adjacent elements. Figure A-18 shows an example of how easily an area of the mesh experiencing divergence problems can be identified; the figure shows water surface elevation and velocity direct ions for all nodes. Areas of physically-ina ccurate water surface elevations will be evident. The cause of the divergence in the solution is the Figure A-18. Model divergence, water surface elevation and velocity vectors

PAGE 234

220 survey data. Most notably along the sloping channel banks, locations submerged for only a small fraction of the year, channel mo rphology does not smoothly vary. Areas of deposition and erosion, incl uding sloughed banks and preferential flow pathways, occur with high frequency. While such features are indicative of field conditions, sudden low or high points can cause divergence with the numer ical solution scheme. By zooming in on these diverging elements and clicking on each node, it is simple to identify these nodes failing to follow the overall morphological tre nd. In these instances adjustment of the elevation of a single node, or even a few adjacent nodes, by less than a centimeter, to better match the surrounding nodes, can solve th e divergence problem. In other instances, elements were poorly shaped, with large interior angles, non-uniform node spacing, or element areas differing from adjacent elements by greater than a factor of two. On a case-by-case basis, nodes were added, delete d, or moved slightly to improve mesh characteristics. Considering the dynamic natu re of the channel morphology and the fact that the survey data itself includes two collection periods, these changes are easily justified as having negligible bearing on the eventual model re sult of interest, river stage at the well cross-section. Model convergence is also a concern, in an effort to minimize computation time. Recall the convergence criteria ar e user defined, and the values require all mesh nodes to demonstrate iteration-to-iteration variability of less than these values for “convergence” to be achieved. Values of 0.1 m2/s and 0.05 meters were used for unit flow (both x and y directions) and water depth convergence, respectively. Pe r the eventual desire to accurately predict river stage, water depth convergence was of particular significance. It was found that when the above convergence cr iteria were satisfied, the average depth

PAGE 235

221 change of all mesh nodes between iterations wa s typically less than one centimeter. This average value, also found in the output file, was perhaps of greater interest, but the FESWMS model does not allow convergence to be specified according to this average change. The repercussion of this model li mitation is that a single diverging node for which unit flow and water depth cannot stab ilize will prevent mode l convergence. The 10 highest nodes can be viewed from the output file. While admittedly, multiple nodes will be outside of the convergence criteria during el ement drying, the limitation is evident. In an attempt to counter this limitation, the top 10 nodes were also recorded for each model simulation. In most instances, less than five nodes would stand out as having significantly larger maximum changes for unit flow and unit depth. When the same nodes top the list for both unit flow and water de pth, these nodes were noted. In the same fashion as warning messages were alleviated elements containing these elements were meticulously studied, and minor elevati on changes or node spacing changes were completed. The act of correcting nodes and elements identified by mesh quality checks, warning messages, convergence criteria listings, or simply looking at the solution is never a guarantee of model success. Mesh refineme nt is always an iterative procedure. Considering the general nature of the wa rning messages or the appearance of an unexpected mesh solution, it is never really assured that the underlying problem has been fixed. Node elevation changes were specifically made to be minimal, in an effort to preserve both the underlying survey data and the relevance of the model solution. In some cases, these adjustments would prove not enough. In some cas es, seemingly fixing one problem would cause a problem with adjacent elements. After each round of

PAGE 236

222 corrections, the model was started again unde r the same boundary conditions. Solutions were analyzed and additional corrections were made. Convergence of the Steady-State Model for Submerged Conditions The first sign of true progress was the attainment of convergence for a simple steady state solution for submerged conditi ons. The second sign of progress was the removal of all warning messages from the mode l output file. Warning messages, after all, are only one step removed from an error messa ge, the latter of which cause the model to crash completely. Nodestrings created within the mesh interior were used as tools to check flow convergence. While these nodestrings are never exactly pe rpendicular to flow direction, they can be used to compute flow though th e cross-section based on nodal velocity and water depth solutions along th e nodestring (integration of flow area and velocity along a line of nodes). Compar ison of flows at the boundary condition nodestrings and within the mesh interior is a useful check that a steady state solution was achieved, given that the model uses alternate criteria for c onvergence). Velocity vectors can also be displayed to check for nodes fa lling outside of the typical range, including nodes with velocity vectors not facing the di rection of flow or node s with unusually large magnitudes for velocity. When a solution is imported into the display window using the Data Browser, several additional statistics are computed and di splayed. Noting the interest in river stage for this research, the range, mean, and standa rd deviation of the water surface elevations are of primary interest. The use of natura l boundary conditions, which allow for modest variability of water surface along the boundary nodestrings, guarantees that the water surface elevation range with in the model solution will exceed the input boundary condition range. Additional variability is the re sult of diverging nodes within the interior

PAGE 237

223 of the mesh, as the water surface should generally slope downward along the model domain. The mean water surface elevation s hould fall close to the average of the boundary conditions at any given time step. Th e water surface standard deviation will be a function of the head gradient through the m odel domain. For the small head gradient of 0.2 meters from upstream to downstream, the standard deviation fo r well-characterized solutions was typically close to 0.035 mete rs. Since node distribution in not uniform within the mesh and the flow path is not stra ight, the statistics were used a guidelines to check model reliability. For dynamic solutions, th ese statistics prove particularly useful as they allow identification of the exact time step at which the solution started to experience divergence difficulties. In these inst ances, the output text files can become too large to open using WordPad. By importing the solution and noting when the problems begin, stepping through each solution time step before a nd after this time enables identification of particular mesh area s in need of further refinement. The above mesh refinement process was re quired to get a simple steady state model to run. No consideration of element drying ha s yet taken place. It is certain that mesh drying will bring about a comple tely new set of problems, some of which have previously been suggested in the discussions pertaining to vertical banks and the establishment of model convergence criteria. Realize, however, that the mesh problems fixed previously would each have likely guaranteed model failure when element drying began. In a sense, these problems were avoided before the star t of what would become the biggest modeling challenge of this project, the simulation of mesh wetting and dr ying and prediction of channel stage over time, as a function of st reamflow. The modified mesh with steady state solutions for velocity (vectors) and wate r depth (color-filled co ntours) are shown in

PAGE 238

224 Figure A-19. Note the smooth nature of the de pth variations from thalweg to vertical banks and the characteristic trend of velocity magnitudes, with highest magnitudes observed for the thalweg and lowe st magnitudes where flow is the shallowest and farthest from the inner channel. The model also in cludes distinct material properties for the channel bed, channel banks, and floodplain ar eas. The channel bed and channel banks are defined by a roughness value of 0.045. The fl oodplain represents th at portion of the model domain that is not submerged during mo st years and to which vegetation imparts a different roughness of 0.10. These values are taken from the RIVMOD input data files. water depth 10.00 10.67 11.33 12.00 12.67 13.33 14.00 14.67 Figure A-19. Converged velocity solution, model domain submerged Dynamic FESWMS Simulations: Lowering of the Water Table Model Parameterization Before testing the ability of the mesh to simulate water surface lowering and element drying, model parameters needed to be set. The ultimate goal is to create a stable model able to simulate in-channel water depths from near zero to bank full conditions. At

PAGE 239

225 times, this effort will require some creativity, but some basic rules of the game need to be laid out. Head boundary conditions will be app lied at the upstream and downstream boundary nodestrings. The hydraulic head loss, or decrease in water surface elevation, along the approximately 500-meter flow path of the model domain will remain constant at 0.2 meters. The rate at which the water surf ace is lowered will vary, but the rate will be the same at each boundary nodestring. Because of the intrinsic instab ilities associated with a jagged wet / dry boundary, boundary conditions are lowered continuously, instead of through the use of a step function wher eby the model is lowered to a point and stabilized to reach a st eady state. At this point, the con cern is not the simulation of actual flow conditions. Rates of head decrease ar e chosen based on two simple criteria: the FLO2DH model being able to run to completi on and the drying of elements appearing to follow the topography of the terra in. Given the potential for rapid stage change within the system, as a result of sudden upstream snowmelt, the model should possess the capability to simulate a rapid lowering of the water surface. The maximum rate of water surface lowering to maintain model stability was f ound to be approximately 0.1 meters per 3.0 hours. In many instances, these same 0.1-me ter decreases required 4 or 5 hours (at minimum) because of the irregular nature of the channel topography. With nodestrings created to include all no des along the upstream and downstream boundaries, boundary conditions are assigned by sele cting ‘FESWMS / Assign BC.’ W ithin this dialog box, the following options are checked: boundary defi nition: specified fl ow / WSE. This highlights for editing the specified flow / WSE options portion of the dialog box. Check the water surface elevation box and the natural boundary circle, the latter to allow small

PAGE 240

226 fluctuations, from the single specified bounda ry condition, in wate r surface elevation along the boundary. For steady state simulations a constant water surface elevation is input. For dynamic simulations, the transient op tion is selected and a series of points are inserted to define the time / water surf ace elevation curve at the selected boundary. Two options are available for starting the m odel, a cold start or a hot start. The FLO2DH model in FESWMS computes wate r surface elevations, water depths, and velocity vectors for all nodes in the model mesh. A cold start is used when no previous FLO2DH solution is available. By default, veloci ty in the x and y directions is zero for all nodes. An initial water surface elevation is assigned to al l nodes not already assigned a water surface elevation by the boundary condi tions. The model will converge sooner the closer this initial value is to the eventual solution. Initial wate r surface elevation is assigned from ‘FESWMS / Model Control / Parameters’ by typing a value in the appropriate box. For a cold start, the init ial water surface must be greater than the elevation of the highest node in the model domain, such that all nodes are submerged. The steady state option is selected unde r ‘FESWMS / Model Control‘, the maximum number of iterations is set to 99, and the model will iterate towards a solution. For the hot start option, the initial conditi ons to the FLO2DH model are read from the output files (.ini and .flo file extensions). Nodes ar e assigned the water surface elevations and velocity vectors from a pr evious simulation. Several advantages are offered by this approach. Steady state si mulations with unchanging boundary conditions can be repeatedly run until meeting the esta blished convergence crite ria, rather than stopping at the solution after 99 iterations. Additionally, dynamic simulations can be run for smaller time periods, and the final solution can be used as the initial conditions for the

PAGE 241

227 next simulation. This keeps the output f ile sizes manageable for post-processing and facilitates early identification of divergence pr oblems. Hot starts can be initialized from any time step of a previous dynamic solution. Th is is particularly us eful in the case of relatively minor divergence pr oblems caused by disjoint wet elements during the drying process. The output .flo file can be edited to force these disjoint elements to be dry (which would be the case in re ality) at the first sign of th eir appearance. By saving this modified file, the model can be restarted fr om this middle time step, and the disjoint nodes will not be carried through future ti me steps. Small divergence problems can therefore be corrected and th e water surface lowering can continue. By contrast, changing any of the node locations or el evations would require starting over again from a cold start with all nodes submerged. In the case of no element divergence due to drying, hot start solutions alleviate having to start over with submerged conditions. This is a time consideration. As noted, an upper limit exists for the rate of water table lowering. If it is known that the model is able to simulate hi gher stage levels, node modification is not a desirable choice. However, large divergence proble ms require node modification, re-starting of the simulation from submerge d conditions, and re-lowering of the water table, with no guarantee that another drying problem will not occur. To use the hot start option, which is available for both steady state and dynamic simulations, the previous solution file is imported using the Data Browse r, and the output file data sets and desired time step to be used as new initial cond itions is selected under ‘FESWMS / Initial Conditions Data.’ Additionally, the .ini external file option must be selected in the FLO2DH input section under ‘FESWMS / Model Control’ and the appropriate .ini file selected.

PAGE 242

228 Model parameters are also set within ‘FESWMS / Model Control.’ Required selections include the following, with the appropriate choice in parentheses: FESWMS version (FESWMS 3), FLO2DH input (.net file a nd .ini file for hot start simulations), run type (hydrodynamic), means to compute botto m stresses (Manning’s Equation), solution type (steady state or dynamic), slip conditions (semi-slip as next discussed), and option for higher order integration (none). The latter two of these parameters have not yet been discussed. The slip option app lies no shear stress (frictionl ess) at the closed boundaries; no slip indicates that the shear stress at closed boundaries is so great that the tangential velocity is zero; and semi-slip allows for s lip at a closed boundary unless the flow is against a vertical wall. Becau se the model boundaries in this model represent vertical walls, most notably at locations of steep, n ear-vertical banks, th e no slip condition is selected to impart a non-zero frictional re sistance to flow. After clicking on the Parameters box, a value of 1.0 is entered for th e vertical wall shear coefficient, signifying that the shear stress on the wa lls is the same as on the bot tom, based on roughness values assigned to elements at the base of the vert ical walls. Higher order integration results in more precise solutions with a lengthier analys is time, primarily suited for dealing with curve-sided elements. Convergence criteria for unit flow and water dept h are also input in this dialog box. Based on results from stea dy state analyses, unit flow convergence equal to 0.1 m2/s and unit depth convergence of 0.05 m were determined to provide reliable results. The element drying / wetting option must be checked if elements dry during a simulation. The timing of element drying is de termined by the input values for storativity and depth tolerance for drying. Different va lues are used dependant on which elements are drying. Storativity values in the range of 0.0 to 0.1 meters were used throughout the

PAGE 243

229 modeling process, with higher values best suited for elements possessing the steepest slopes. Depth tolerance for drying values in cluded a minimum of 0.1 meters for stage levels on the channel bed, 0.3 meters for stag e levels along the channel banks, and values of 2.0 meters when flows approach the inner channel thalweg. These values are necessary to prevent excessive oscillation of elements between wet and dry, while not preventing re-wetting altogether. All other parameters in FESWMS remain ed unchanged from the default values. This conclusion is based on a review of the physical significance of the default values (with the SMS Help menu as a reference) and an assessment of data limitations. Default values are used for the average water densit y and momentum flux co rrection coefficients. Model defaults also exclude consideration of Coriolis forces and wind conditions. Within the iterations box under Model Cont rol, a relaxation factor of 0.01, as recommended by other model users, is us ed to increase the numerical stability. Presumably due to the result of the high degree of channel variability, the FLO2DH model will crash when the default value of 1.0 is used. A time step of 15 minutes (0.25 hours) was specified. Maximum iterations we re set at 30 for dynamic simulations. Element drying often prevented model convergence based on the convergence parameters, but trials indicated that this numbe r of iterations resulted in a stable solution. The time integration factor was not changed from the default value of 0.667. The start time and end time values are in reference to transient boundary conditions assigned under dynamic simulations. Material properties are assigned for both floodplain and channel bank elements. Manning roughness values are taken from the RI VMOD input data files, with 0.10 used

PAGE 244

230 for floodplain elements and 0.045 used for elem ents within the confines of the main channel. Turbulence parameters, including eddy viscosity and eddy diffu sivity remain at the default values. Model Operation Element wetting and drying: considerations and parameterization To predict channel stage as a function of flow using SMS, the model must be capable of simulating element drying wit hout crashing. Althou gh the FESWMS model does include options to simulate element dryi ng, the ability of the model to converge is dependant on the pattern in which elements dry. It is preferable for elements to dry as the layers of an onion peel, always with a smoot h wet / dry boundary, as might be reasonably expected for a straight, sy mmetric channel with smooth elevation changes throughout. Unfortunately, these optimal characteristics do not describe the Carson River mesh. Cross-sectional area is variable, including both smooth (gradually sloping) and abrupt (vertical banks) transition zones from channe l bed to floodplain. In the downstream half of the modeled reach, channel thalwegs ar e deeper and the inner channel elements transport a significantly greater fraction of to tal flow than upstream, where thalwegs are less incised and more discrete. The modeled re ach is not straight, including an almost 90 curve in the downstream section. Channe l morphology is not smooth, including small hills and valleys where otherwise the trends in elevation change are reversed. Portions of the channel bed include se dimentation deposits and er osional deposits from bank sloughing. With the mesh constructed from a network of survey points with irregular node spacing and from multiple time periods, some of these bed features are captured better than others.

PAGE 245

231 These observations dictate the element drying pattern when the water surface is gradually lowered. Elements cannot dry al ong a smooth boundary. Higher areas of the bank and channel bed will become dry, unable to transport flow, before elements at lower elevations. The rows of quadrilateral elements simply cannot be constructed to follow the contours of the land. Irregular drying result s in element re-wetting and the potential for wet elements that are disconnected from the ma in channel. The oft-jagged nature of the wet / dry boundary causes numerical diverg ence at corner node s of the outermost elements. As higher elements internal to the outermost wet / dry boundary become dry, flow can no longer be transported through th ese elements. Effectively, water flowing from left to right is suddenly faced w ith flowing around an obstacle, and numerical divergence often results. Node s internal to the active bound ary nodestrings also dry in these situations, potentially impairing the ab ility of the model to transport flow. Drying along the boundary can result in supercritical flow conditions that can be managed by the FLO2DH model. However, supercritical flow conditions internal to the model mesh cause a model crash. The convergence criteria have pr eviously been discussed. The values set are strict enough that the model will never truly “converge” if the wet / dry boundary is jagged in several locations within the mesh. Rather, the model will perform the maximum number of iterations before moving forward. For this exercise, convergence qualities of the mesh were measured in an alternate fashion based on both the statistics for each imported solution and a measure of common sense. With the exception of nodes located at wet / dry element interfaces, along with natural fluctuation along the boundary nodestrings, the water surface elevations for nodes internal to the model mesh s hould fall in the range

PAGE 246

232 between the head boundary conditions Each imported solution is also visually inspected. Water surface elevations not in accord with neighbor ing nodes and velocity vectors not in line with the flow direction or possessing uncharacteristic magnitudes were noted. While such divergence is expected along the wet / dry boundary, such divergence within the submerged portion of the mesh requires correc tive action to be taken. In many instances, this divergence will manifest itself by causing element re-wetting, disjoint wet elements within the mesh, and model fa ilure. Within a dynamic soluti on, model divergence can be identified when the standard deviation of the water surface elevations in the solution deviates significantly from 0.035 meters, under the 0.2-meter hydraulic gradient from the boundary conditions. Model divergence: node identificati on, causes, and solution examples The ultimate goal is to develop a model able to continuously lower the water surface elevation from above bank-full to n early zero without the model crashing or diverging to a physically-unr easonable solution. The rate s of change in the head boundary conditions are allowed to vary, as ne eded to best handle drying of each channel feature, but the rate remain s constant for both boundary conditions for each time step. In other words, the 0.2-meter hydraulic gradient is a fixed value. The process here is iterative. The model must initialize with all node s submerged. The water surface is then lowered by 0.1 meters over a period of be tween 3 and 6 hours. One cause of model failure is “water surface < 0 at node X in element Y.” When drying along the steeper elements along the rip-rap and gradually-slo ping channel banks, it is not desirable for these elements to dry when only one node b ecomes dry, given that the elements do not perfectly follow the land contours, and a stora tivity value of 0.1 meters is used to allow

PAGE 247

233 the elements to continue to transport flow. The 0.1-meter va lue is nothing more than an estimated average of elevation loss across these elements. By the time the computed water surface has fallen to 0.1 meters below the first dry node, the element is most likely transporting very little flow and allowed to become dry. Mo del failure occurs for less steep elements when the computed water surface falls below all element nodes, but not more than 0.1 meters below any one node. In many instances, this effect is caused by computational instabilities and can be rectified by simply drying these elements at a slower or faster rate, by a ltering the boundary condition time series. Elements may also re-wet (depth tolerance for drying = 0.3 meters for bank elements) and become disjoint because of similar computationa l instabilities associated w ith the 15-minute time steps. Smaller time steps are also unable to remedy this problem. Additionally, note that by altering the rate of water surf ace lowering closer to steady st ate values, the same end is achieved; the rate of change (change in boundary condition divided by time step) is the same in both cases. If the model crashes unde r this scenario, which is evident when looking at a dynamic solution, changing the rate of change is also an option worthy of attempt. Due to the flatter nature of elements within the channel bed, a storativity value of 0.0 meters is used, but this problem does still present itself for nodes along the downstream boundary as the model attempts to force all nodes to closely match the designated head value. Unfortunately, most model crashes and solution inconsistencies were linked to specific diverging nodes. In these instances, the problem is typically due to problems with the node spacing and elem ent size, node elevations th at do not follow the general elevation change trend, or both. To aid with visualization, pi cture a line of elements, with

PAGE 248

234 therefore a line of nodes al ong the tops and bottoms of these elements. Divergence problems arise when the elevations along a line of nodes is rising or close to steady, but within that line are one or tw o or three nodes representing an elevation dip. Similarly, sudden high points occurring along the dow nward sloping channel bed can cause divergence problems. Of course, such natura l variability is not a by-product of poor survey techniques, but rather this simply captures the variability present in the Carson River system. When nodes are cl osely spaced together, as they are with the 3 to 5 meter typical element length in the Carson River me sh, the finite difference algorithm may be unable to compensate. Sudden elevation cha nges on the order of only a few centimeters can pose divergence problems. The highest dens ity of these divergen t nodes occurs along the channel banks and rip-rap because of the slower velocities, where fluid momentum is less of an influencing factor. While at first these divergent nodes ma y not be obvious in a solution, such problems become evident afte r error propagation thr ough several iterations and time steps. Divergent nodes may not cause immediat e model failure, but error propagation does lead to divergence problems with adjacent nodes and an unreasonable model solution. Nodes diverging for these reasons are generally ea sily corrected by changing one or two node elevations by a few centim eters, although generally less than three centimeters. Such modest changes in the elev ation of even two hundred nodes within this 12,000+ node model will negligibly alter the final solution. The changes certainly fall within the error alr eady associated with assuming a c onstant channel bathymetry for modeling the streamflow, predicting stage, a nd facilitating a comparison with water table levels in the groundwater wells.

PAGE 249

235 When a divergence problem is identified fr om either a visual inspection of the solution, solution statistics, or a model cr ash, the number of each diverging node is recorded. The output .prt file is generated after all model runs, regardless of whether a model crash occurs. This file includes the assigned model parameters; initial conditions; convergence checks; and solution water surface elevation, water depth, and velocity vector data for all 12,000+ nodes in the model mesh. Node solution data is available for each iteration of a steady state run or for any time step interval of a dynamic solution, for which the 15-minute time step was selected. Notably, the FLO2DH model can be parameterized to print out up to the top 10 nodes for the convergence check component, ranked based on value change since the previ ous iteration or time step. By setting this parameter to 10, the 10 nodes with the maximum changes in y-direction unit flow, x-direction unit flow, and wate r surface elevation can be viewed for each time step of the model solution. The average change in these values for all model node s is also noted. For a given time step, either the final time step be fore a model crash or a time step at which solution divergence is identified as beginning, those nodes with the clearest convergence problems were recorded. Particular interest was given to the nodes with the maximum changes with regard to water surface elevation. If an area of the mesh is seen that consistently produces the larges t changes, it is a good idea to refine the elements in this portion of the mesh to assist with model convergence. The highest ranked nodes often also appeared near the tops of the unit fl ow convergence check lists, and multiple nodes from the same element were not uncommon. W ith a list of divergi ng nodes in hand, the process of mesh editing can begin. Node el evations for these and adjacent nodes were modified as needed to aid model stability Nodes were also moved to improve the

PAGE 250

236 uniformity of node spacing and element shap e. Nodes along the wet / dry boundary were generally ignored, as divergen ce is due simply to there location along this boundary and the associated numerical instability. Lastly mesh quality was checked and any problems were corrected. The mesh was then saved under a new name. One might inquire as to the reasoning behi nd editing nodes that have yet to show a definite divergence problem. Personal e xperience with the FESWMS FLO2DH model indicates that these nodes will eventually caus e a model problem after sufficient time for error propagation. In addition, after a mesh edit, the model must be re-initialized with all nodes submerged. A series of steady state simulations are completed with all nodes submerged until the convergence criteria ar e met. With no element drying and mesh corrections already completed for submer ged conditions, the model satisfies the convergence criteria, starting from a cold st art, within three st eady state runs of 99 iterations each and using previous runs as hot start files. The water surface is then lowered to the levels at which divergence or a crash occurred (with the final steady state solution as a hot start file), so as to find out if the corrections were adequate to fix the problem. If the changes were too conservati ve, the mesh editing, the steady state runs, and the water surface lowering will need to be repeated. The water surface elevation can not be instantaneously lowered to the desire d test levels, making th is process extremely time consuming. Presumably, if all was done properly, the model will not crash or diverge before this lower stage level is reached or surpassed. Mesh editing was performed iteratively in this fashion until a lower limit was found at which the model would work properly. As more and more experience was gained with FLO2DH, some mesh problems were identifi ed and corrected in advance of a model

PAGE 251

237 crash or discovery of a divergence problem The mesh layout prior to water surface lowering and additional mesh editing is shown again in Figure A-20. While more than 20 of these editing iterations were performed, the mesh editing procedure and problems encountered are as previously discussed. elevation 1250.00 1250.67 1251.33 1252.00 1252.67 1253.33 1254.00 1254.67 Figure A-20. Mesh structure before me sh editing to improve element drying After considerable editing, the FLO2DH hydrodynamic model is able to simulate element drying from bank-full to the point of transition from channel banks to channel bed. The modeling experience and insight ga ined from fixing the many problems was next applied to the elements along the channel bed. Additional elements were added along the downstream portion of the thalwegs to prevent too many large elements from drying within a given time step. Element density was also increased for the slope transition from bank elements to bed elements. Efforts were made to create bed elements of a fairly uniform size and rectangular shap e. Triangular elements were included only Problem Area 8 Problem Area 1 Problem Area 2 Problem Area 3 Problem Area 4 Problem Area 5 Problem Area 6 Problem Area 7

PAGE 252

238 where needed around channel bends and changes in cross-sectional area, with one notable exception. Quadrilateral elements were split into triangular elements in those downstream areas in which the thalweg becomes branch ed. These changes assist the model in accurately simulating the channel flow under low-flow conditions. The drying of elements internal to the main body of flow due to depositional bed features along the channel bed, requires careful mesh editing to prevent model divergence or failure when flow runs into these dynamically-changing obstacl es to flow. Mesh quality was also used to correct gradient and element area problems before they could cause model divergence. Due to the flatter slope along th e channel bed, except the thal weg, a storativity value of 0.0 meters and a depth tolerance for drying value of 0.1 meters were used. For flows passing primarily through thalweg elements, stor ativity values remained at 0.0 and depth tolerance was increased to 2.0 meters. This e ffectively prevents element re-wetting, but it is necessary to prevent elements from oscill ating between wet and dry due the steep side slopes.

PAGE 253

239 APPENDIX B NEAR-CHANNEL WELL RESPONSE TO STREAMFLOW VARIABILITY 0 200 400 600 800 1000 1200 1400 1600 1800 2000 3/1/20013/4/20013/7/20013/10/20013/13/20013/16/ 20013/19/20013/22/20013/25/20013/28/20013/31/2001 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-1. March 2001

PAGE 254

240 0 50 100 150 200 250 300 350 400 450 500 550 600 3/22/20013/25/20013/28/20013/31/2001 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-2. March 22-21, 2001 0 200 400 600 800 1000 1200 1400 1600 1800 2000 4/1/20014/4/20014/7/20014/10/20014/13/20014/16/ 20014/19/20014/22/20014/25/20014/28/20015/1/2001 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-3. April 2002

PAGE 255

241 0 100 200 300 400 500 600 700 800 900 1000 4/21/20014/24/20014/27/20014/30/2001 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-4. April 21-30, 2001 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 5/1/20015/4/20015/7/20015/10/2001 date and timeCarson River Streamflow (cfs)1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-5. May 1-9, 2001

PAGE 256

242 0 200 400 600 800 1000 1200 1400 1600 1800 2000 11/1/200111/4/200111/7/200111/10/200111/13/200111/ 16/200111/19/200111/22/200111/25/200111/28/200112/1/2001 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-6. November 2001 0 50 100 150 200 250 300 11/24/200111/25/200111/26/200111/27/2001 date and timeCarson River Streamflow (cfs)1251.5 1251.6 1251.7 1251.8 1251.9 1252.0 1252.1 1252.2Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-7. November 24-27, 2001

PAGE 257

243 0 200 400 600 800 1000 1200 1400 1600 1800 2000 12/1/200112/4/200112/7/200112/10/200112/13/200112/ 16/200112/19/200112/22/200112 /25/200112/28/200112/31/2001 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-8. December 2001 0 50 100 150 200 250 300 12/3/200112/4/200112/5/200112/6/200112/7/2001 date and timeCarson River Streamflow (cfs)1251.7 1251.8 1251.9 1252.0Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-9. December 3-6, 2001

PAGE 258

244 0 200 400 600 800 1000 1200 1400 1600 1800 2000 1/1/20021/4/20021/7/20021/10/20021/13/20021/16/ 20021/19/20021/22/20021/25/20021/28/20021/31/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-10. January 2002 0 50 100 150 200 250 1/18/20021/19/20021/20/20021/21/20021/22/2002 date and timeCarson River Streamflow (cfs)1251.7 1251.8 1251.9 1252.0 1252.1 1252.2Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-11. January18-21, 2002

PAGE 259

245 0 50 100 150 200 250 300 1/29/20021/30/20021/31/20022/1/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-12. January 29-31, 2002 0 200 400 600 800 1000 1200 1400 1600 1800 2000 2/1/20022/4/20022/7/20022/10/20022/13/20022/ 16/20022/19/20022/22/20022/25/20022/28/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-13. February 2002

PAGE 260

246 0 50 100 150 200 250 2/1/20022/2/20022/3/20022/4/2002 date and timeCarson River Streamflow (cfs)1251.6 1251.7 1251.8 1251.9 1252.0 1252.1 1252.2 1252.3Water Table Elavation, NAVD 88 (meters) Scout Camp UBC Well 1 Water Level Figure B-14. February 1-3, 2002 0 200 400 600 800 1000 1200 1400 1600 1800 2000 3/1/20023/4/20023/7/20023/10/20023/13/20023/16/2002 3/19/20023/22/20023/25/20023/28/20023/31/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-15. March 2002

PAGE 261

247 0 200 400 600 800 1000 1200 1400 1600 1800 2000 4/1/20024/4/20024/7/20024/10/20024/13/20024/16/ 20024/19/20024/22/20024/25/20024/28/20025/1/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-16. Near-channel well response to streamflow variability, April 2002 0 100 200 300 400 500 600 4/1/20024/2/20024/3/20024/4/20024/5/20024/6/20024/7/20024/8/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-17. April 1-7, 2002

PAGE 262

248 0 100 200 300 400 500 600 700 800 900 1000 4/8/20024/9/20024/10/20024/11/20024/12/20024/13/20024/14/20024/15/2002 date and timeCarson River Streamflow (cfs)1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-18. April 8-14, 2002 0 200 400 600 800 1000 1200 1400 1600 1800 4/15/20024/16/20024/17/20024/18/2002 date and timeCarson River Streamflow (cfs)1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8 1252.9Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-19. April 15-17, 2002

PAGE 263

249 0 200 400 600 800 1000 1200 1400 1600 1800 2000 5/1/20025/4/20025/7/20025/10/20025/13/20025/16/ 20025/19/20025/22/20025/25/20025/28/20025/31/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-20. May 2002 0 100 200 300 400 500 600 5/1/20025/2/20025/3/20025/4/20025/5/20025/6/20025/7/20025/8/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-21. May 1-7, 2002

PAGE 264

250 0 100 200 300 400 500 600 700 800 5/8/20025/9/20025/10/20025/11/20025/12/20025/13/20025/14/20025/15/2002 date and timeCarson River Streamflow (cfs)1251.8 1251.9 1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-22. May 8-14, 2002 0 200 400 600 800 1000 1200 1400 1600 1800 5/15/20025/16/20025/17/20025/18/20025/19/20025/20/20025/21/20025/22/2002 date and timeCarson River Streamflow (cfs)1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-23. May 15-21, 2002

PAGE 265

251 0 200 400 600 800 1000 1200 1400 1600 1800 5/25/20025/26/20025/27/20025/28/20025/29/20025/30/20025/31/20026/1/2002 date and timeCarson River Streamflow (cfs)1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-24. May 25-31, 2002 0 200 400 600 800 1000 1200 1400 1600 1800 2000 6/1/20026/4/20026/7/20026/10/20026/13/20026/16/ 20026/19/20026/22/20026/25/20026/28/20027/1/2002 date and timeCarson River Streamflow (cfs)1250.0 1250.5 1251.0 1251.5 1252.0 1252.5 1253.0 1253.5Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-25. June 2002

PAGE 266

252 0 200 400 600 800 1000 1200 1400 1600 6/1/20026/2/20026/3/20026/4/2002 6/5/20026/6/20026/7/20026/8/2002 date and timeCarson River Streamflow (cfs)1252.0 1252.1 1252.2 1252.3 1252.4 1252.5 1252.6 1252.7 1252.8Water Table Elavation, NAVD 88 (meters) Fort Churchill Gage Scout Camp UBC Well 1 Water Level Figure B-26. June 1-7, 2002

PAGE 267

253 APPENDIX C ELEMENT DRYING AND WATER SURFAC E LOWERING WITHIN COMPLETED SMS MODEL water depth : 46.000 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-1. Northwest channel bank

PAGE 268

254 water depth : 64.500 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-2. Southwest channel bank rip-rap water depth : 76.000 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-3. North-central channel bank

PAGE 269

255 water depth : 85.000 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-4. Northeast channel bank water depth : 115.000 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-5. Southeast channel bank

PAGE 270

256 water depth : 123.000 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-6. Western channel bed water depth : 140.500 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-7. Central channel bed

PAGE 271

257 water depth : 143.500 0.00 0.40 0.80 1.20 1.60 2.00 2.40 2.80 3.20 3.60 4.00 Figure C-8. Channel bed, minimum fl ow level for 1.99-meter head loss

PAGE 272

258 LIST OF REFERENCES Ambrose, R.B., Wool, T.A ., Martin, J.P. and Schanz, R.W., 1991. WASP5.X—A Hydrodynamic and Water Quality Model: Model theory, User’s Manual, and Programmer’s Guide. U.S. EPA, Athens, Georgia. Amyot, M., Kraepiel, A.M.L. and Morel, F.M.M., 1998. The Chemical Cycle and Bioaccumulation of Mercury. Annual Revi ews: Ecology and Systematics, 29, 543566. Arteaga, F.E., 1986. Mathematical Model Analysis of the Eagle Valley Ground-Water Basin, West-Central Nevada. Nevada Divi sion of Water Resources Bulletin 45. Babiarz, C.L., Hurley, J.P., Benoit, J.M., Shafer, M.M., Andren, A. W. and Webb, D.A., 1998. Seasonal Influences on Partitioning and Transport of Total and Methylmercury in Rivers from Contras ting Watersheds. Biogeochemistry, 41, 237257. Balogh, S.J., Meyer, M.L. and Johnson, D.K ., 1997. Mercury and Suspended Sediment Loadings in the Lower Minnesota River. Environmental Science and Technology, 31(1), 198-202. Barlow, P.M., DeSimone, L.A. and Moench, A.F., 2000. Aquifer Response to Streamstage and Recharge Variations: II. Convol ution Method and Applications. Journal of Hydrology, 230, 211-229. Bloom, N.S. and Effler, S.W., 1990. Seasonal Variability in the Me rcury Speciation of Onondaga Lake (New York). Water, Air, and Soil Pollution, 53(3-4), 251-265. Bonzongo, J.C., Heim, K.J., Chen, Y., Lyons, W.B., Warwick, J.J., Miller, G.C. and Lechler, P.J., 1996a. Mercury Pathways in the Carson River-Lahontan Reservoir System, Nevada, USA. Environmental T oxicology and Chemistry, 15(5), 677-683. Bonzongo, J.C., Heim, K.J., Warwick, J.J. and Lyons, W.B., 1996b. Mercury Levels in Surface Waters of the Carson River-Lahontan Reservoir System, Nevada: Influence of Historic Mining Activities. Envi ronmental Pollution, 92(2), 193-201. Boulton, A.J., Findlay, S., Marmonier, P., Stanley, E.H. and Valett, H.M., 1998. The Functional Significance of the Hyporheic Zone in Streams and Rivers. Annual Reviews: Ecology and Systematics, 29, 59-81.

PAGE 273

259 Callister, S.M. and Winfrey, M.R., 1986. Micr obial Methylation of Mercury in Upper Wisconsin River Sediments. Water, Ai r, and Soil Pollution, 29(4), 453-465. Carroll, R.W.H., 1999. Simulation of Mercury Transport and Fate in the Carson River, Nevada. Masters Thesis. University of Nevada-Reno, Reno, Nevada. Carroll, R.W.H., James, A.I., Warwick, J.J. and Miller, J.R., in pr ess. Modeling Erosion and Overbank Deposition During Extreme Flood Conditions on the Carson River, Nevada. Journal of Hydrology. Carroll, R.W.H. and Warwick, J.J., 2001. Un certainty analysis of the Carson River mercury transport model. Ecological Modeling, 137, 211-224. Carroll, R.W.H., Warwick, J.J., Heim, K.J ., Bonzongo, J.C., Miller, J.R. and Lyons, W.B., 2000. Simulation of Mercury Trans port and Fate in the Carson River, Nevada. Ecological Modeling, 125, 255-278. Chen, Y., Bonzongo, J.C., Lyons, W.B. and Miller, G.C., 1997. Inhibition of Mercury Methylation in Anoxic Freshwater Sedi ment by Group VI Anions. Environmental Toxicology and Chemistry, 16(8), 1568-1574. Chen, Y., Bonzongo, J.C. and Miller, G.C., 1996. Levels of Methylmercury and Controlling Factors in Surface Sediment s of the Carson River System, Nevada. Environmental Pollution, 92(3), 281-287. Cooper, H.H., Jr. and Rorabaugh, M.I., 1963. Changes in Ground-Water Movement and Bank Storage Caused by Flood Waves in Surf ace Streams. U.S. Geological Survey Professional Paper 475-B, B192-B195. Covay, K.J., Banks, J.M., Bevans, H.E. and Watkins, S.A., 1996. Environmental and Hydrologic settings of the Las Vegas Valley Area and the Carson and Truckee River Basins, Nevada and California. U.S. Geological Survey Water-Resources Investigations Report 96-4087. de Magalhaes, M.E. and Tubino, M., 1995. A Possible Path for Mercury in Biological Systems: The oxidation of Metallic Me rcury by Molecular Oxygen in Aqueous Solutions. The Science of the Total Environment, 170(3), 229-239. Dever, R.J., Jr. and Cleary, R.W., 1979. Un steady-State, Two-dimensional Response of Leaky Aquifers to Stream Stage Fluctuat ions. Advances in Water Research, 2, 1318. Fetter, C.W., 2001. Applied Hydrogeology. Pren tice Hall, Inc., Upper Saddle River, New Jersey. Fitzgerald, W.F. and Clarkson, T.W., 1991. Mercury and Monomethylmercury: Present and Future Concerns, Global Atmos pheric Change and Human Health. Environmental Health Perspectives, 96, 159-166.

PAGE 274

260 Fitzgerald, W.F. and Mason, R.P., 1997. Bi ogeochemical Cycling of Mercury in the Marine Environment. In: A. Sigel and H. Sigel (editors), Merc ury and its Effects on Environment and Biology-Metal Ions in Bi ological Systems. Marcel Dekker, Inc., pp. 53-111. Freeze, R.A., 1971. Three-dimensional, Transi ent, Saturated-Unsaturated Flow in a Groundwater Basin. Water Resour ces Research, 7(2), 347-366. Freeze, R.A. and Cherry, J.A., 1979. Groundwater Prentice-Hall, Inc, Englewood Cliffs, New Jersey. Froehlich, D.C., 2000. Finite Elemen t Surface-Water Modeling System: TwoDimensional Flow in a Horizontal Pl ane, Version 2, User’s Manual. Fuller, C.C. and Harvey, J.W., 2000. Reactive Uptake of Trace Metals in the Hyporheic Zone of a Mining-Contaminated Stream Pinal Creek, Arizona. Environmental Science and Technology, 34(7), 1150-1155. Gill, G.A. and Bruland, K.W., 1990. Mercur y Speciation in Surface Freshwater Systems in California and Other Areas. Environm ental Science and Technology, 24(9), 1392-1400. Gill, M.A., 1985. Bank Storage Characteristics of a Finite Aquifer Due to Sudden Rise and Fall of River Level. Jour nal of Hydrology, 76, 133-142. Gilmour, C.C. and Henry, E.A., 1991. Mercur y Methylation in Aquatic Systems Affected by Acid Deposition. Environmental Pollution, 71, 131. Gilmour, C.C., Henry, E.A. and Mitchell, R., 1992. Sulfate Stimulation of Mercury Methylation in Freshwater Sediments. Environmental Science and Technology, 26(11), 2281-2287. Govindaraju, R.S. and Koelliker, J.K., 1994. Applicability of Lin earized Boussinesq Equation for Modeling Bank Storage under Un certain Aquifer Parameters. Journal of Hydrology, 157, 349-366. Hall, F.R. and Moench, A.F., 1972. Applicat ion of the Convolution Equation to Stream Aquifer Relations. Water Res ources Research, 8(2), 487-493. Hantush, M.S., 1967. Flow of Groundwater in Relatively Thick Leaky Aquifers. Water Resources Research, 3, 583. Hantush, M.M., Harada, M. and Marino, M.A ., 2001. Modification of Stream Flow Routing for Bank Storage. EPA/600/A-01/052. Harada, M., Hantush, M.M. and Marino, M.A ., 2000. Hydraulic Analysis on StreamAquifer Interaction by Storage F unction Models. EPA/600/A-00/071.

PAGE 275

261 Heim, K.J. and Warwick, J.J., 1997 Simulati ng Sediment Transport in the Carson River and Lahontan Reservoir, Nevada, USA. J ournal of the American Water Resources Association, 33(1), 177-191. Hoffman, R.J. and Thomas, K.A., 2000. Methylmercury in Water and Bottom Sediment Along the Carson River System, Nevada and California, September 1998. U.S. Geological Survey Water-Resources Investigations Report 00-4013. Hornberger, G.M., Ebert, J. and Remson, I., 1970. Numerical Solution of the Boussinesq Equation for Aquifer-Stream Interaction. Water Resources Research, 6(2), 601-608. Hosseinipour, E.Z. and Martin, J. L., 1990. RIVMOD—A One-dimensional Hydrodynamic and Sediment Transport Mode l: Model Theory and User’s Manual. Hudson, J.M., Gherini, S.A., Watras, C.J. and Porcella, D.B., 1994. Modeling the Biogeochemical Cycle of Me rcury in Lakes: The Merc ury Cycling Model (MCM) and its Application to the MTL Study Lakes. In: Watras, C.J. and Huckabee, J.W. (editors), Mercury Pollution: Integrati on and Synthesis. Lewis Publishers, Boca Raton, New York, pp. 473-523. Hunt, B., 1990. An Approximation for the Bank Storage Effect. Water Resources Research, 26(11), 2769-2775. Hurley, J.P., Benoit, J.M., Babiarz, C.L., Shafer, M.M., Andren, A.W., Sullivan, J.R., Hammond, R. and Webb, D.A., 1995. Influe nces of Watershed Characteristics on Mercury Levels in Wisconsin Rivers. Environmental Science and Technology, 29(7), 1867-1875. Jones, J.B., Jr. and Holmes, R.M., 1996. Surf ace-Subsurface Interactions in Stream Ecosystems. Trends in Ecology and Evolution (TREE), 11(6), 239-242. Kennedy, E.J., 1984. Discharge Ratings at Gaging Stations. Techniques of WaterResources Investigations of the United States Geological Survey, Chapter A10. Lechler, P.J., Miller, J.R., Hsu, L. and Desilets, M.O., 1997. Merc ury Mobility at the Carson River Superfund Site, West-Centr al Nevada, USA: Interpretation of Mercury Speciation Data in Mill Tailings Soils, and Sediments. Journal of Geochemical Exploration, 58, 259-267. Lin, H.J., Richards, D.R., Talbot, C.A.,Yeh, G., Cheng, J., Cheng, H. and Jones, N.L., 1996. FEMWATER: A Three-Dimensional Fi nite Element Computer Model for Simulating Density-Dependant Flow and Tr ansport in Variably Saturated Media, Version 3.0. Marino, M.A., 1975. Digital Simulation Mode l of Aquifer Response to Stream Stage Fluctuation. Journal of Hydrology, 25, 51-58.

PAGE 276

262 Martin, J.L., 1992. MERC4: A Mercury Transport and Kinetics Model: Model Theory and User’s Guide. U.S. EPA, Athens, Georgia. Marvin-DiPasquale, M. and Oremland, R.S ., 1999. Methylmercury Formation and Degradation in Sediments of the Cars on River System, Interim Report I. U.S Geologic Survey Water Resources Division, Menlo Park, CA. Matilainen, T., 1995. Involvement of Bacteria in Methylme rcury Formation in Anaerobic Lake Waters. Water, Air, a nd Soil Pollution, 81(1-4), 757-764. Maurer, D.K., 1986. Geohydrology and Simu lated Response to Ground-water Pumpage in Carson Valley, a River-dominated Basin in Douglas County, Nevada and Alpine County, California. U.S. Geological Surv ey Water-Resources Investigations Report 86-4328. Meade, R.H., Moody, J.A. and Stevens, J.H.H., 1995. Contaminants in the Mississippi River 1987-1992. U.S. Geological Survey Circular 1133. Miller, J., Barr, R., Grow, D., Lechler, P., Richardson, D., Waltman, K. and Warwick, J., 1999. Effects of the 1997 Flood on the Transport and Storage of Sediment and Mercury within the Carson River Valley, West-Central Nevada. The Journal of Geology, 107, 313-327. Miller, J.R. and Lechler, P.J., 1998. Mercury Partitioning within A lluvial Sediments of the Carson River Valley, Nevada: Implicati ons for Sampling Strategies an Tropical Environments. In: J.W.e. al. (editor), Geochemistry of Tropical Environments. Springer-Verlag, pp. 211-233. Miller, J.R., Lechler, P.J. and Desilets, M., 1998. The Role of Gepmorphic Processes in the Transport and Fate of Mercury in the Carson River Basin, West-Central Nevada. Environmental Geology, 33(4), 249-262. Miller, J.R., Rowland, J., Lecher, P.J., De silets, M. and Hsu, L., 1996. Dispersal of Mercury-Contaminated Sediments by Ge omorphic Processes, Sixmile Canyon, Nevada, USA: Implications to Site Char acterization and Reme diation of Fluvial Environments. Water, Air, and Soil Pollution, 86, 373-388. Moench, A.F. and Barlow, P.F., 2000. Aquife r Response to Stream-stage and Recharge Variations: I. Analytical Step-response Functions. J ournal of Hydrology, 230, 192210. Moench, A.F., Sauer, V.B. and Jennings, M.E., 1974. Modification of Routed Streamflow by Channel Loss and Base Fl ow. Water Resources Research, 10(5), 963-968. Morel, F.M.M. and Hering, J.G., 1993. Prin ciples and Applications of Aquatic Chemistry. John Wiley and Sons, New York, NY.

PAGE 277

263 Morel-Seytoux, H.J., 1975. A Combined Model of Water Table and River Stage Evolution. Water Resources Research, 11(6), 968-972. Neuman, S.P., 1975. Analysis of pumping test data from anisotropic unconfined aquifers considering delayed gravity response. Water Resources Research, 11, 329-342. Neuman, S.P. and Witherspoon, P.A., 1969. Appli cability of current th eories of flow in leaky aquifers. Water Resources Research, 5, 817-829. Neuman, S.P. and Witherspoon, P.A., 1971. An alysis of Nonsteady Flow with a Free Surface Using the Finite Element Method. Water Resources Research, 7(3), 611623. Nevada Division of Water Planning (N DWP), 1996. Carson River Chronology. Perkins, S.P. and Koussis, A.D., 1996. Stream -Aquifer Interaction Model with Diffusive Wave Routing. Journal of H ydraulic Engineering, 210-218. Pinder, G.F. and Sauer, S.P., 1971. Numeri cal Solution of Flood Wave Modification Due to Bank Storage Effects. Water Resources Research, 7(1), 63-70. Porvari, P. and Verta, M., 1994. Meth ylmercury Production in Flooded Soils: A Laboratory Study. Water, Air, and Soil Pollution, 80(1-4), 765-773. Schaefer, D.H. and Whitney, R., 1992. Geological Framework and Ground-water Conditions in Basin-fill Aquifers of the Dayton Valley and Churchill Valley Hydrographic Areas, Western Nevada. U. S. Geological Survey Water-Resources Investigations Report 91-4072. Serrano, S.E. and Workman, S.R., 1998. M odeling Transient Stream/Aquifer Interaction with the Non-linear Boussinesq Equation a nd its Analytical So lution. Journal of Hydrology, 206, 245-255. Sharp, J.M., Jr., 1977. Limitations of BankStorage Model Assumptions. Journal of Hydrology, 35, 31-47. SMS: Surface Water Modeling System, Tuto rial: Version 8.0, 2000. Brigham Young University – Environmental Mo deling Research Laboratory. Squillace, P.J., 1996. Observed and Simula ted Movement of Bank-Storage Water. Ground Water, 34(1), 121-134. Stordal, M.C. and Gill, G.A., 1995. Determin ation of Mercury Methylation Rates using a 203-Hg Radiotracer Technique. Water, Ai r, and Soil Pollution, 80(1-4), 725-734. Thomann, R.V.and Mueller, J.A., 1987. Principles of Surface Water Quality Monitoring and Control. Harper Collins Publishers, New York, New York, pp. 549.

PAGE 278

264 Tindall, J.A. and Kunkel, J.R., 1999. Unsatu rated Zone Hydrology for Scientists and Engineers. Prentice-Hall, Inc., U pper Saddle River, New Jersey. Trexler, D.T., 1977. Geologic Map, Carson C ity Quadrangle. Nevada Bureau of Mines and Geology Urban Map Series, Carson C ity Folio, Map 1Ag, scale 1:24000. United States Department of Health and Human Services (USDHHS), 1997. Toxicological Profile for Mercury. Public Health Service. United States Environmental Protection Ag ency (USEPA), 1997. Mercury Study Report to Congress, Volume III: Fate and Trans port of Mercury in the Environment. EPA452/R-97-003. USEPA, 2000. Mercury Research Strategy. EPA/600/R-00/073. United States Geological Survey (USGS), 1995. Mercury Contamination of Aquatic Ecosystems. U.S. Geological Survey Fact Sheet 216-95. USGS, 1997. Flood of January 1997 in the Cars on River Basin, Ca lifornia and Nevada. U.S. Geological Survey Fact Sheet 183-97. van Rijn, L.C., 1984a. Sediment Transport, Pa rt I: Bed Load Tran sport. Journal of Hydraulic Engineering, 110(10), 1431-1456. van Rijn, L.C., 1984b. Sediment Transport, Part II: Suspended Load Transport. Journal of Hydraulic Engineering, 110(11), 1613-1641. Warwick, J.J. and Heim, K.J., 1995. Hydrodynamic Modeling of the Carson River and Lahontan Reservoir, Nevada. Water Resources Bulletin, 31(1), 67-77. Watras, C.J., Bloom, N.S, Claas, S.A., Morri son, K.A, Golmour, C.C. and Craig, S.R., 1995a. Methylmercury Production in the Anoxic Hypolimnion of a Dimictic Seepage Lake. Water, Air, and Soil Pollution, 80(1-4), 735-745. Watras, C.J., Morrison, K.A., Host, J.S. and Bloom, N.S., 1995b. Concentration of Mercury Species in Relationship to othe r Site-Specific Factors in the Surface Waters of Northern Wisconsin Lakes. Limnology and Oceanography, 40(3), 556565. Wayne, D.M., Warwick, J.J., Lechler, P.J., Gill, G.A. and Lyons, W.B., 1996. Mercury Contamination in the Carson River, Nevada : A Preliminary Study of the Impact of Mining Wastes. Water, Air, and Soil Pollution, 92, 391-408. Whiting, P.J. and Pomeranets, M., 1997. A Numerical Study of Ba nk Storage and its Contribution to Streamflow. Jour nal of Hydrology, 202, 121-136.

PAGE 279

265 Wise, W.R., Clement, T.P. and Molz, F. J., 1994. Variably Saturated Modeling of Transient Drainage: Sensitivity to Soil Pr operties. Journal of Hydrology, 161, 91108. Workman, S.R., Serrano, S.E. and Liberty, K ., 1997. Development and Application of an Analytical Model of Stream/Aquifer Inter action. Journal of Hydrology, 200, 149163. Xun, L., Campbell, N.E.R. and Rudd, J.W .M., 1987. Measurements of Specific Rates of Net Methylmercury Production in the Wa ter Column and Surface Sediments of Acidified and Circumneutral Lakes. Cana dian Journal of Fi sheries and Aquatic Sciences, 44, 750-757. Yeh, W.W-G., 1970. Nonsteady Flow to a Surface Reservoir. ASCE Journal of Hydraulics, 96(3), 609-618. Zillioux, E.J., Porcella, D.B. and Benoit, J.M., 1993. Mercury Cycling and Effects in Freshwater Wetland Ecosyste ms. Environmental Toxicology and Chemistry, 12, 2245-2264. Zitta, V.L. and Wiggert, J.M., 1971. Flood Rout ing in Channels with Bank Seepage. Water Resources Research, 7(5), 1341-1345. Zlotnik, V.A. and Huang, H., 1999. Effect of Shallow Penetration and Streambed Sediments on Aquifer Response to Stream Stage Fluctuations (Analytical Model). Ground Water, 37(4), 599-605.

PAGE 280

266 BIOGRAPHICAL SKETCH Daniel Eugene Crawford was born in Jack sonville, Florida on February 22, 1978. Raised in Jacksonville, he departed to attend the University of Florid a in the Fall of 1996. This was the best of times to be a Gator football fan. He graduated with a B.S. in environmental engineering in the Fall of 2001. Staying on for another 2 years, he completed the requirements for his M.E. in e nvironmental engineering in the Spring of 2003.


Permanent Link: http://ufdc.ufl.edu/UFE0000856/00001

Material Information

Title: Simulation of Spatial and Temporal Variability of Methylmercury Concentration within Channel Banks and Surface Waters of the Carson River, Nevada
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000856:00001

Permanent Link: http://ufdc.ufl.edu/UFE0000856/00001

Material Information

Title: Simulation of Spatial and Temporal Variability of Methylmercury Concentration within Channel Banks and Surface Waters of the Carson River, Nevada
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0000856:00001


This item has the following downloads:


Full Text












SIMULATION OF SPATIAL AND TEMPORAL VARIABILITY OF
METHYLMERCURY CONCENTRATION WITHIN CHANNEL BANKS AND
SURFACE WATERS OF THE CARSON RIVER, NEVADA















By

DANIEL E. CRAWFORD


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF ENGINEERING

UNIVERSITY OF FLORIDA


2003

































Copyright 2003

by

Daniel E. Crawford

































This thesis is dedicated to the graduate students of the University of Florida.















ACKNOWLEDGMENTS

I would like to thank the National Science Foundation for their generous support of

this project through grant # 4510586-12. I wish to acknowledge the time, effort, and

continued support of my advisor John J. Warwick. The well-being of his students always

remained a priority. Special thanks go to Andy James for his support and guidance

throughout this research, most notably during the 7 months after John's departure. I also

thank Bill Wise and Jean-Claude Bonzongo for serving on my committee and providing

invaluable feedback. I thank my peers and closest friends for helping me (through

countless hours of frustration) with words of encouragement; and for always being there.

Lastly, I would like to thank my parents and family. Although the research may not have

been their forte, their support, encouragement, and sense of pride in my accomplishments

never waned. My time at this university has been long-lived, but the experience was

greater than ever expected and has truly prepared me for the future.
















TABLE OF CONTENTS

page

A C K N O W L ED G M EN T S ............................ ........ ....................................................... iv

LIST OF TABLES ...... ..... ...... ........ ....... .. ......... ......... .... viii

LIST OF FIGURES ......... ........................................... ............ ix

AB STRA CT.................................................... ............................. xiii

CHAPTER

1 RESEARCH LOCATION: AN INTRODUCTION TO THE CARSON RIVER........ 1

C arson R iver B asin ....... .. .. ...... .. ...... .......................... .. ................ ..
C arson R iver Stream fl ow ............................................................. ....................... 2
Carson River Aquifer System s .................................. .....................................5
H historical M ining O operations ............................................... ............................ 5
Carson R iver G eom orphology ......................................................... ................ ..... 7

2 MERCURY CONTAMINATION AND MERCURY CYCLING ............................10

Mercury Cycling within Surface Water Systems .................................... ............10
Quantification of Mercury Levels in the Carson River System ...............................11
Introduction of Mercury to the Carson River System ...........................................13
Chemistry of M ercury in Soil ................................. ............... ............... 15
M ercury Transform ation and Toxicity ............................................ ............... 16
M ethylm ercury in the W ater Colum n..................................................................... 19
M ercury in the Aquatic Food Chain ................................ ......................... ....... 23
Carson R iver Superfund Site .................................. ................ ............... 27
Health Effects of M ercury Contamination ........................ .... ......... .............. 28

3 PREVIOUS MODEL STRUCTURE ............................... .........................30

Hydrodynamic Modeling Using RIVMOD.....................................................30
The W ASPS and M ERC4 Program s...................................... ......................... 33
Stages of M odel D evelopm ent...................................................................... ... ..... 36









4 CHANNEL BANK ELEMENTS: LINKING SURFACE WATER AND
G R O U N D W A T E R ......... .. ....... .............. .................................................. 38

M odeling the C arson R iver.............................................................. .....................38
A available D ata ........................................40
Introduction to the H yporheic Zone ......................... ....... ....... ............... .... 41
Modeling Surface-Subsurface Interactions: A Historical Perspective .....................43
M modeling Sim plifications ........................................ ... ................................. 50
Using IFMTFM Modifications to Track Bank Saturation: Bank Element
Functionality ............................................... ................. ........... 50
Final Considerations and Comparative Analysis............................................ 60

5 SURFACE W ATER M ODELING ..................................... ......................... ......... 63

Site Selection .......................... .....................63
Collection of GPS Survey Data ........... .. .............................. 64
Surface Water Model Selection ........... ...... ........ ................... 66
Functionality of the FESWMS Model......................................... ...............67
Creation of Scatter Point Data Set.................................................................. 70
Pre-Processing of Data Prior to M esh Construction .......................... ...................70
Mesh Construction and Editing as Discussed in Appendix A..................................72
Sum m ary of M esh E diting .......... ... ........ .... ..... .................. .....................74

6 WATER TABLE RESPONSE TO STREAMFLOW VARIABILITY ....... ........ 77

Linking Surface Water Modeling and Groundwater Data ............... .....................77
Groundwater W ell Data................................................ ....................85
Quantifying the Relationship between River Stage and Water Table Elevation........95

7 BANK ELEMENT DEVELOPMENT AND COMPARISON TO PREVIOUS
IFM TFM ............... ................................... ........................... 137

General Considerations for Bank Element Design and Operations........................137
Updated Input Data Sets to Include Bank Elements ............ ............................... 139
Comparison of Updated IFMTFM Code versus Previous Version ................... 144
N ext Step for the IFM TFM M odel ........................................ ....................... 164

8 COMPARATIVE ANALYSIS: RIVMOD VERSUS FESWMS MODULE OF
S M S ............................................................................... 1 6 7

Comparison of Channel Geometry along the Well Transect................................169
Steady-State Comparisons for Velocity and Depths of Flow................................169
Comparative Analysis and Data Limitations................. ...................... ...............176









APPENDIX

A SMS Model Parameterization and Mesh Editing ............................................. 179

B Water Table Response to Streamflow Oscillations ...............................................239

C Element Drying and Water Surface Lowering within Completed SMS Model .......253

L IST O F R E FE R E N C E S ............................................... ............................ ................258

B IO G R A PH IC A L SK E T C H ........................................ ............................................266
















LIST OF TABLES


Table page

4-1. Spatial variability of reaction constants within IFMTFM.......................................53

6-1. Classification of surveyed water surface elevations according to time of
collection ..................................... ................................. ......... 109

6-2. Steady-state output for depth and discharge from final SMS model ...................124

7-1. Spatial variability of M/D reaction constants within IFMTFM .............................146

7-2. Segment classification for M/D reaction constants in 'old' IFMTFM................. 156

7-3. Segment classification for M/D reaction constants in 'new' IFMTFM...............56

7-4. Error associated with sample M eHg data points...................................................161

7-5. Parameters unchanged between 'old' and 'new' IFMTFM versions.....................161

7-6. Comparison of channel bank MeHg modeling in 'old' and 'new' IFMTFM ........161

8-1. Comparison of steady-state velocity and depth output from SMS and
RIVM OD ..................................... ................................. .......... 171
















LIST OF FIGURES


Figure page

1-1. Map of the Carson River drainage and selected sampling sites..............................3

1-2. General locations of known mill tailings deposits and Scout Camp field site...........7

2-1. T he aquatic m mercury cycle .......................................................................... ..... 11

2-2. Processes that transport mercury into the water column along the Carson River.... 18

3-1. M odified channel geometry ........................................................................32

4-1. Water table response to changing river stage............... ................. ............... 50

4-2. State variables measured with MERC4 ........................................ ...............54

4-3. Dependency of methylmercury concentration on soil moisture.............................. 57

5-1. Spatial location of survey points and cross-sections at Scout Camp site ...............65

5-2. Original survey points after import into SMS scatter module...............................72

5-3. Bathymetry of Scout Camp reach after mesh editing found in Appendix A ..........75

5-4. Minimum flow levels able to be modeled, subject to a 0.2-meter head gradient.....76

6-1. 2001 USGS streamflow gage comparison: 15-minute data...................................81

6-2. 2002 USGS streamflow gage comparison: 15-minute data .....................................81

6-3. May 2002 Carson River streamflow at USGS gages ............................................82

6-4. Complete water table data from Well 1 and Well 3 at Scout Camp site ...............87

6-5. Transient water levels in Well 1 and Well 3, 2001-2002...................................89

6-6. Hydraulic differential from Well 1 to Well 3, 2001-2002 ....................................90

6-7. Pump test drawdown data collected for Well 1 in October 2002...........................91









6-8. 2001 streamflow routing comparison: FCH gage data and RIVMOD simulation
for upstream boundary condition ........................................ ......................... 96

6-9. 2002 streamflow routing comparison: FCH gage data and RIVMOD simulation
for upstream boundary condition ........................................ ......................... 97

6-10. April 2002 streamflow routing comparison: FCH gage data and RIVMOD
simulation for upstream boundary condition ................................. ............... 97

6-11. 2001: near-channel well response to streamflow variability............................... 99

6-12. 2002: near-channel response to streamflow variability .............. ... ................100

6-13. April 2002: near-channel response to streamflow variability .............................101

6-14. April 1-7, 2002: near-channel response to streamflow variability .........................101

6-15. April 8-14, 2002: near-channel response to streamflow variability .......................102

6-16. April 15-17, 2002: near-channel response to streamflow variability ...................102

6-17. May 2002: near-channel response to streamflow variability...............................103

6-18. May 1-7, 2002: near-channel response to streamflow variability ........................103

6-19. May 8-14, 2002: near-channel response to streamflow variability ......................104

6-20. May 15-21, 2002: near-channel response to streamflow variability ....................104

6-21. May 25-31, 2002: near-channel response to streamflow variabililty .....................105

6-22. February 2002: near-channel response to streamflow variability ........................105

6-23. February 1-3, 2002: near-channel response to streamflow variability .................106

6-24. Hydrodynamic routing through RIVMOD segments within reach of interest.......108

6-25. Water surface elevation (WSE) as a function of flow for March 2001 data ..........109

6-26. March 9, 2002: location of surveyed water surface observations in RIVMOD
se g m en t 1 5 2 ...................................................................... 1 1 1

6-27. March 9, 2001: graphical approximation of channel water surface slope in
RIVM OD segm ent 152 ..................................................... ..................... 111

6-28. Zero time lag: temporal correlation between river stage and water table
response in W ell 1 ................................... ............................. ...... 114









6-29. Zero time lag, flow spikes removed: temporal correlation between river stage
and water table response in W ell 1............... ............. .................116

6-30. One hour time lag: temporal correlation between river stage and water table
response in W ell 1 ........ ........... ........... .... ................. ............... 117

6-31. Two hour time lag: temporal correlation between river stage and water table
response in W ell 1 ...................... ........ .......... ..... ...................... 117

6-32. Channel thalweg location based on survey points from March and July, 2001 .....121

6-33. Linear regression of March, July 2001 surveyed bed elevations for use in
determination of hydraulic gradient for SMS model ............... ...................121

6-34. Stage Discharge Relationship for Steady State Lowering of SMS Water
Surface E elevation ....................... .. .... .................................... .. ............ .. 125

6-35. Linear regression for determination of discharge as a function of depth ..............127

6-36. Linear regression for determination of depth as a function of discharge .............129

6-37. 2001 rating curve-based estimates for water surface elevation in the channel
com pared to water table level in W ell 1..................................... ............... 131

6-38. 2002 rating curve-based estimates for water surface elevation in the channel
com pared to water table level in W ell 1..................................... ............... 131

7-1. Station and segment definition within IFMTFM ............................................. 140

7-2. Conceptual IFMTFM segment interaction..................................142

7-3. Relationship between temporally-averaged RIVMOD stage and bank
satu ration h eight .................................................................................. 150

7-4. Medium flow (May 16, 1994) IFMTFM calibration for inorganic Hg(II)
(parameter ki) in channel banks for 'old' and 'new' code versions ....................153

7-5. High flow (June 10, 1995) IFMTFM verification for inorganic Hg(II)
(parameter ki) in channel banks for 'old' and 'new' code versions ....................154

7-6. Low flow (June 16, 1994) IFMTFM verification for inorganic Hg(II)
(parameter 1i) in channel banks for 'old' and 'new' code versions ....................154

7-7. Medium flow (May 16, 1994) IFMTFM calibration for MeHg (k2) in channel
banks ('old' code version) and K20 methylation rate in channel banks ('new'
co d e v ersio n ) ..................................................................... 15 8









7-8. High flow (June 10, 1995) IFMTFM calibration for MeHg threshold height
(HT) in channel banks ('old code version) and verification for MeHg in
channel banks ('new code version)........................................ ......................... 158

7-9. Low flow (June 16, 1994) IFMTFM verification for MeHg in channel banks
for 'old' and 'new code versions ............................................... ............... 159

7-10. IFMTFM simulation for May / June 1995 flood event at Fort Churchill (model
segment 140), illustrating MeHg dilution effect of threshold height HT in 'old'
co d e v version ...............16................. .........3................................. 1

7-11. IFMTFM simulation for May / June 1994 flows at Fort Churchill (model
segment 140), demonstrating time lag for MeHg response in 'new' code
version ..................................... ................................. ........... 163

8-1. Comparison of channel geometry at well transect for SMS and RIVMOD........... 170

8-2. Q = 109.6 cms: comparison of steady-state velocity output from SMS and
RIVM OD ..................................... ................................. ......... 173

8-3. Q = 6.85 cms: comparison of steady-state velocity output from SMS and
RIVM OD ..................................... ................................. ......... 173

8-4. Q = 109.6 cms: comparison of steady-state depth predictions from SMS and
RIVM OD ..................................... ................................. ......... 175

8-5. Q = 6.85 cms: comparison of steady-state depth predictions from SMS and
RIVM OD ..................................... ................................. ......... 175

8-6. Comparison of contacted bank height predictions from SMS and RIVMOD .......176















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Engineering

SIMULATION OF SPATIAL AND TEMPORAL VARIABILITY OF
METHYLMERCURY CONCENTRATION WITHIN CHANNEL BANKS AND
SURFACE WATERS OF THE CARSON RIVER, NEVADA

By

Daniel E. Crawford

August 2003

Chair: John J. Warwick
Major Department: Environmental Engineering Sciences

The USEPA designated the Carson River, Nevada as part of a Superfund site in

1990. Between 1860 and 1890, elemental mercury was used at over 70 mill sites to

recover gold and silver via mercury amalgamation, introducing 7000 tons of mercury to

the Carson River watershed. Contaminated sediments line channel banks downstream of

these mill sites, with mercury concentrations two to three orders of magnitude above

regional background levels. Mercury concentrations in both the Carson River and

Lahontan Reservoir are among the highest known worldwide. Mercury methylation

within sediments (accompanied by diffusion, advection, and erosion processes)

introduces toxic methylmercury to the water column for ultimate transport into Lahontan

Reservoir. Methylmercury concentrations in the bank are not spatially and temporally

constant; but are instead a function of moisture history.

A computer simulation model, Integrated Fluvial Mercury Transport and Fate

Model (IFMTFM), was developed to simulate bank erosion and diffusion from bottom









sediments. A new bank element was developed and incorporated into the previous

IFMTFM, to relate flow history within the channel to moisture-content levels observed

within the channel banks. Zones of varying moisture affect methylation and

demethylation kinetics; and therefore affect temporal concentrations of methylmercury

available to the water column.

Global positioning system (GPS) data were used to construct a detailed, temporal

model of three-dimensional river hydrodynamics for a specific reach of the model

domain, using Surface Water Modeling System (SMS) software. Streamflow and water

table data show a tight coupling between the Carson River and the surficial aquifer. The

nearly instantaneous water table response to streamflow oscillations (as observed in a

monitoring well set back 5 feet from the channel bank) serves as the foundation for bank

element functionality.

Incorporation of the bank element into the complete Carson River model of

hydrodynamics, sediment transport, and mercury transport and cycling (IFMTFM)

provides a new capability to simulate temporally variable mercury concentrations within

the channel bank; reduces the number of model calibration parameters from three to two;

and increases the accuracy of in-stream mercury predictions. The height of bank

saturation and zone of active microbial methylation are continuously updated as river

stage changes. Contacted bank sediments release mercury to the water column through

erosion, advection, and diffusion processes; with no threshold height for methylation.

Hydrodynamic components of IFMTFM (RIVMOD) and SMS were compared to show

the value of high-resolution transect data.














CHAPTER 1
RESEARCH LOCATION: AN INTRODUCTION TO THE CARSON RIVER

Carson River Basin

The Carson River, located in the southwestern corner of Nevada, drains from the

eastern Sierra Nevada Mountains just south of the Lake Tahoe Basin (NDWP, 1996). The

Sierra Nevada Mountains thus serve as the river basin's primary source of precipitation.

Precipitation in the Carson River Basin is attenuated by the rainshadow of the Sierra

Nevada. In the modeled portion of the Carson River (downstream of Carson City, Figure

1-1), the climate is semi-arid, with typical annual precipitation of less than 5 inches

(Covay et al., 1996). Vegetation near the river is largely sagebrush or similar species,

with cottonwood trees growing along the riparian zone immediately adjacent to the river.

The Carson River Basin encompasses an area of approximately 3,966 square miles

in the states of California and Nevada (Figure 1-1). The basin is topographically closed.

A closed basin has no outlet, and water entering such a basin discharges only by

evapotranspiration or loss to groundwater (Covay et al., 1996). The upper portion of the

Carson River Basin is drained by the Carson River's East and West Forks. The Carson

River's two forks merge in the northern part of Carson Valley and form the Carson River

mainstem, which then continues on toward the river system's terminus in the Carson

Sink. The Carson Valley agricultural area diverts water from both river forks and from

the Carson River mainstem, to provide irrigation for approximately 35,000 acres.

Lahontan Reservoir, completed in 1915 with the construction of Lahontan Dam, serves as

a major sediment trap upstream of the Carson Sink. The Truckee Canal diverts water









from the Truckee River Basin into the reservoir, to meet the agricultural water demands

associated with the Newlands Irrigation Project (downstream of Lahontan Dam). United

States Geological Survey (USGS) streamflow gages are available at Carson City (CCG

station #10311000), Fort Churchill (FCH #10312000), Lahontan Dam (LDM

#10312150), and the Truckee Canal (TRC #10351400).

Of the Carson River Basin's total surface area, approximately 606 square miles

(387,840 acres; or just over 15%) lie within the State of California. The remaining 3,360

square miles (2,149,680 acres; or almost 85%) lie within the State of Nevada (NDWP,

1996). While most of the basin's surface area is in the State of Nevada, most of the

basin's precipitation originates in the State of California. The geographic separation

between the Carson River Basin's principal supply of water and its principal demands has

historically resulted in controversy surrounding the rights to and uses of water resources.

The lower Carson River Basin begins at the head of Dayton Valley and ends at the

furthermost extent of the Carson Sink (NDWP, 1996). The geology of this region

(Lahontan Valley), the bottom of a prehistoric lake, is characterized by sediment deposits

between 3,000 and 8,000 feet in thickness. The present-day Lahontan Valley and Carson

Desert, now considerably drier, is characterized by a number of important and varied

hydrologic features including lakes, marshes and wetlands, cultivated farmlands and

pastures, barren desert sinks, alkali flats, and playas.

Carson River Streamflow

The rain-shadow effect of the Sierra Nevada Mountains to the west allows little

precipitation to fall within the Lahontan Valley (NDWP, 1996). Average annual

precipitation is only about 5 inches, while annual potential evaporation exceeds 60

inches.

































Boundary


1 Carson City gage (CCG)
2 Deer Run Road (DRR)
3- Dayton (DAY)
4 Upstream Table Mountain (UTM)
5 Fort Churchill (FCH)


N 10 0 10 20
S\ KIometers


Figure 1-1. Map of the Carson River drainage and selected sampling sites. (Reprinted
with permission from Carroll, R.W.H., 1999. Simulation of Mercury
Transport and Fate in the Carson River, Nevada. Masters Thesis. University
ofNevada-Reno, Reno, Nevada. Figure 1.1, pp. 9)









Before extensive upstream storage and water diversions, the Lahontan Valley was

generally inundated in late winter and early spring by the unhindered runoff of the

melting snow pack from the Sierra Nevada Range. After this runoff period, a drying

pattern would ensue during the remainder of the year until the winter precipitation period

commenced again in October-November. This highly seasonal pattern of wet and dry has

been disrupted since the completion of the Newlands Project in the valley in 1903, and

particularly after the completion of Lahontan Reservoir in 1915.

Carson River flow is primarily from snowmelt, but the effects of upstream storage

and storage in Lahontan Reservoir have tended to moderate flow magnitudes. The cycle

of seasonal high flows in the spring (due to snowmelt) and low flows during the

remainder of the year in the Carson River upstream of the Lahontan Reservoir continues,

but the flow magnitudes no longer reflect pre-development conditions. Snowmelt,

whether immediate or from upstream reservoir releases, is the source of Carson River

streamflow; this observation, along with the semi-arid Nevada climate and the extensive

diversion of water for agricultural use along the entire Carson River, results in a highly

variable streamflow pattern.

Carson River flows may vary by three to four orders of magnitude on an annual

basis. During late summer and early fall, streamflow in downstream reaches of the

Carson River may be nonexistent. By contrast, consider the rain-on-snow flood event of

January 1997; the Carson River extended well into the floodplain, with peak discharge

values of 24,000 cubic feet per second (cfs) and 22,300 cfs estimated at Carson City

Gage (CCG) and Fort Churchill Gage (FCH), respectively (USGS, 1997).









Carson River Aquifer Systems

Principal aquifers in the Carson River Basin are unconsolidated basin-fill deposits

of interbedded gravel, sand, silt, and clay (Covay et al., 1996). Aquifers within the basin

readily receive, transmit, and store large volumes of water. Detailed hydrologic

investigations of the Carson River Basin have been performed as part of the USGS

National Water-Quality Assessment (NAWQA) program. General areas along the Carson

River underlain by basin-fill aquifers include Carson Valley, Eagle Valley, Dayton

Valley, and Churchill Valley.

Unconsolidated deposits occur in all valleys of the Carson River Basin and are the

primary water-bearing units (Covay et al., 1996). Carson Valley contains unconsolidated

Quaternary basin-fill sediments (mostly fine- to coarse-grained Carson River alluvium),

and semiconsolidated Tertiary sediments (mostly lake and stream deposits) (Maurer,

1986). Eagle Valley contains Quaternary semiconsolidated and unconsolidated alluvium

(including deposits of sand, silt, and coarse materials; and clay confining units) (Trexler,

1977; Arteaga, 1986). Basin-fill deposits in Dayton and Churchill Valleys include

Cenozoic sedimentary deposits (500 to 3000 ft thick); alluvial fans and pediments; and

lake, river, delta, and floodplain deposits (consisting of a mixture of clays, silts, sands,

and gravel) (Schaefer and Whitney, 1992). The sediments within the modeled reach of

the Carson River Valley clearly may be defined as typical alluvium.

Historical Mining Operations

Historical mining activities have contaminated lower reaches of the Carson River,

Lahontan Reservoir, and Stillwater Marsh with mercury. Extensive milling and ore

processing operations were situated within Dayton Valley during the era of the Comstock

Lode during the late 1800s (NDWP, 1996). Here the Carson River passes through a









canyon and loses 240 feet of elevation over a distance of 10 miles. With its proximity to

the Comstock mines to the north and the available hydropower, this was an ideal location

to establish the mills. During the peak of the Comstock era, more than 75 mills operated

along the Carson River.

Gold Canyon and Sixmile Canyon (Figure 1-2) mark the geologic boundaries of

one of this nation's richest silver strikes-the Comstock Lode. Known mill tailings

deposits are found in areas along the Carson River from just downstream of Carson City

to Dayton and in areas along the Carson River tributaries from Virginia City (Wayne et

al., 1996), as shown by the crosshatched areas in Figure 1-2. "Efforts to extract this vast

body of silver wealth would begin a period of natural resource exploitation (timber

cutting and water diversions) and environmental degradation (deforestation, soil erosion,

water pollution, arsenic and mercury spills), the effects of which have carried forward to

the present day" (NDWP, 1996). During the Comstock's peak years of production (1862

to 1880), approximately $296,400,000 worth of silver and gold were removed from the

mines. Over its entire active mining period, which lasted from 1857 to 1921, the

Comstock yielded $386,346,931 (historical value) in total mineral production (of which

$222,315,814 was in silver; $164,023,917 was gold; and $7,200 was copper and lead).

Mining operations associated with the Comstock Lode resulted in the production of

an extensive volume of mine tailings (waste rock material). It has been estimated that

during some peak periods of mining, more than 600 tons of tailings material were

released to the surrounding area on a daily basis (Miller et al., 1998). Liquid elemental

mercury was used to extract gold and silver from the mined material using an

amalgamation process. This introduced an estimated 7000 tons of mercury into the















































Figure 1-2. General locations of known mill tailings deposits and Scout Camp field site

Carson River system through mercury residual within the mine tailings (Carroll, 1999),

along with the immense, immeasurable volume of associated sediment. At the mills,

mined ores were pulverized and mixed with mercury, forming a mercury-silver/gold

amalgam because of the high affinity of mercury for these metals (a method known as the

Washoe process). Float tanks were used to separate the amalgam from the waste rock.









Roasting the amalgam, releasing the mercury from the complex as a vapor, would then

produce a pure form of the precious metals. Although most of the mercury vapor was

recovered, it was estimated that as much as 0.68 kg of mercury was released to the

environment with mill tailings for every ton of ore that was processed (Miller et al.,

1996). During the past 140 years, mercury-contaminated sediments were eroded from the

mill sites, transported downstream by fluvial processes, and re-deposited in specific

localities, thereby expanding the influence of mercury pollution (Miller et al., 1996).

Carson River Geomorphology

Analysis of historical documents and alluvial stratigraphy indicate that the Carson

River underwent significant changes in channel form and process during the period of the

Comstock Lode (Miller et al., 1998). The increased sediment load resulting from the

constant supply of mine tailings, along with sediment introduced from clear-cut

headwater areas used for timber production during this period, caused the river to aggrade

(on the order of several meters); and to experience lateral instability and meander

abandonment. The valley floor and abandoned channels subsequently filled with

contaminated debris. The end result of these processes was the storage of

mercury-enriched sediments within a complexly structured alluvial sequence located all

along the Carson River valley downstream of Carson City (Miller et al., 1999). Mercury

concentrations in the alluvial valley fill deposits exhibit great spatial variability

(longitudinally and laterally along the river; and vertically within sediment deposits), but

the concentrations are generally two to three orders of magnitude greater than regional

background levels of 0.01 to 0.10 [tg/L (Miller et al., 1998).









The eventual reduction of mining operations decreased the sediment load and

caused the river to downcut through these contaminated sediments. Today, the Carson

River continues to reclaim its original floodplain in an attempt to establish a new

equilibrium. Mercury-contaminated sediment is believed to enter the system primarily

during higher flow events when water levels reach overlying contaminated sediments

(notably within the floodplain) and erosional processes become significant. Initially,

mercury released from mines represented a point-source contaminant problem. Modern

mercury pollution of the Carson River must be considered a distributed-source problem

because of the contaminated sediment stored within channel banks.














CHAPTER 2
MERCURY CONTAMINATION AND MERCURY CYCLING

Mercury Cycling within Surface Water Systems

Mercury is introduced to surface waters through atmospheric deposition, surface

runoff (erosion), and diffusion. After mercury enters into surface water, it enters a

complex cycle of transformations-primarily according to redox chemistry and microbes.

A schematic drawing of mercury cycling in an aquatic ecosystem is shown in Figure 2-1

(USGS, 1995). Mercury may remain in the water column or be lost through drainage or

advection. Mercury can be incorporated into the bed sediments by particle settling,

perhaps only to be later released by diffusion or resuspension. The formation of

methylmercury can introduce mercury into the food chain, or volatilization may exile the

mercury back to the atmosphere. Divalent mercury and methylmercury, critical

components of the Carson River system, can be transported to water bodies in runoff,

typically bound to suspended soil/humus or attached to DOC (dissolved organic carbon),

or Hg(II) and methylmercury can leach into the water body from groundwater flow in the

upper soil layers (USEPA, 1997). Research has shown that both DOC and pH can play a

significant role in the fate of mercury within a system. Studies have shown that DOC

concentration and pH levels have a strong effect on mercury fate, with increasing acidity

(decreasing pH) and increasing DOC content resulting in higher body burdens in fish

(USGS, 1995). The movement of mercury will vary within each given body of water.

Mercury concentrations are typically reported for the water column (dissolved or










particulate), the underlying sediment (surface or deep sediments), and biota (generally

fish) (USEPA, 1997).


AQUATIC MERCURY CYCLE
| DEPOSITION DEPOSITION

ND DEPOSITiON AND DEPOSITION


h ETP. o LA r. f-. '"C. DEPOSII.ON




yll '* lBIOMAGNIFICATION -' .
.** SDIMENTATION DIFFUSIO SEDIMENTATiON
SEOiME#MT
m "RESUSPENSiON






Figure 2-1. The aquatic mercury cycle. (USGS, 1995. Mercury Contamination of
Aquatic Ecosystems. U.S. Geological Survey Fact Sheet 216-95.
Figure 6, pp. 3)


Quantification of Mercury Levels in the Carson River System

The major water issue for the lower Carson River Basin focuses not on quantity but

rather quality (NDWP, 1996; Carroll, 1999; Miller et al., 1998). Mercury is introduced

into the river from the contaminated mine tailings distributed throughout the riverbank

sediments and floodplain deposits. Mercury concentrations in the tailings are two to three

orders of magnitude greater than background concentrations in the region. Sampling by

the United States Environmental Protection Agency (USEPA) found mercury

concentrations as high as 4,600 parts per million (ppm) in the fine-grained materials of

the tailings piles. Mercury dispersal from these tailings is extensive. Concentrations

exceeding 100 ppm have been measured more than 60 kilometers (km) downstream of









the mining operations. During the most intense period of mining activity, the Carson

River emptied directly onto an extensive wetland now known as the Stillwater Wildlife

Management Area (SWMA). The completion of Lahontan Dam in 1915 created a new

settling basin for sediments derived from the mercury-contaminated mine tailings, but a

large amount of mercury had already been introduced to the wetland habitat downstream

of the dam.

Several extensive site-characterization studies addressed mercury contamination

along the Carson River System. Total mercury (Hg) in three mountain streams indicates

background levels of < 2 nanograms per liter (ng/L) for this region of Nevada (Wayne et

al., 1996). Water samples collected in 1992 and 1993 showed the following:

* Elevated (e.g., > 20 ng/L) mercury levels first appear downstream from
accumulations of mill tailings.
* Total mercury concentrations in unfiltered and filtered water increase downstream.
* Mercury concentrations in both the Carson River and Lahontan Reservoir are among
the highest known worldwide (100 to 1000 ng/L); filtered water samples suggest that
the >0.4 |tm particle fraction constitutes over 60% of the total water-borne mercury
(Wayne et al., 1996).

Abnormally high levels of Hg, Au, and Ag occur in sediments downstream of

former ore milling operations. Systematic trends are apparent from the analysis of bank

and bed samples (Wayne et al., 1996): mine tailings have the highest Hg, Au, and Ag

contents, and fine-grained silts and muds from the river banks typically have higher Hg

contents than coarser sands and gravel from the river channel; and mercury contents in

both channel and bank deposits increase with increasing distance from CCG. Potential

explanations for the high mercury levels in the Carson River system when compared to

other contaminated sites include the large quantity of elemental mercury introduced to the









system and the time period of over 100 years allowed for physical dispersal and chemical

alteration (Wayne et al., 1996).

Introduction of Mercury to the Carson River System

The original elemental form of mercury released to the environment remains the

prevalent form in highly contaminated soils, sediments, and tailings (Lechler et al.,

1997). Subaerial (buried) tailings typically contain greater than 95% elemental mercury;

downstream of the influence from the tailings, along the Carson River, the percentage of

total mercury as elemental mercury increases to slightly greater than 90% before

decreasing further downstream (Lechler et al., 1997). Elemental mercury possesses a high

volatility and low water solubility, and equilibrium conditions of exposed sediments does

result in both volatilization and some dissolution. The species of primary interest in terms

of mercury toxicity are the organo-mercury compounds such as methylmercury (CH3Hg

or MeHg) (Wayne et al., 1996). However, methylmercury is the end product of the

methylation of inorganic divalent mercury (Hg(II) or Hg2+). Elemental mercury (Hg(0) or

Hgo) is the mercury form introduced to the Carson River System. The low reactivity of

the prevalent Hg(0) species does not allow for direct methylation of this species.

Therefore, the environmental transformations of elemental mercury are next considered.

Starting with elemental mercury, the two potential pathways for mercury oxidation

involve volatilization and dissolution. Following its introduction and transport in the

atmosphere (after volatilization), Hg(0) (gas) can undergo transformation to Hg(II) (gas)

and Hg(II) (particulate) within cloud water (USEPA, 1997). This mechanism has been

proposed as an explanation for the global nature of mercury pollution, as it can result in

mercury deposition to land and water. The circulation of air masses on the global scale

has the potential to take Hg(0) emissions from their point of origin and carry them to









anywhere on the globe before transformation and deposition occur. The slow oxidation of

Hg(0) to Hg(II) results in Hg(0) having an atmospheric residence time on the order of 1

year. Other forms of mercury are likely deposited before they are ever able to dilute

within the global atmosphere. The end result of these atmospheric transformations is that

mercury released to the atmosphere deposits mainly as Hg(II), whether from the direct

deposition of emitted Hg(II) or transformation of emitted Hg(0) to Hg(II) by the above

process (USEPA, 1997). Elevated deposition rates therefore occur around emission

sources, but regional or global transport and deposition also occurs. In other words, even

though the principal emissions of mercury are from point sources concentrated in

industrial areas, mercury pollution is truly a global concern.

While this is a pathway that does occur, the arid climate, and hence lack of rainfall,

of the Carson River region and the likelihood of wider atmospheric transport make it an

unlikely source for the large-scale introduction of Hg(II) to this system. Rather, a small

fraction of the liquid Hg(0) that is associated with the mill tailings is dissolved out of the

amalgam particles and subsequently adsorbed to fine-grained sediments. These sediments

are then deposited in downstream, low-energy reaches of the river and reservoir (Lechler

et al., 1997). Research by de Magalhaes and Tubino (1995) has shown that Hg(0) can be

easily oxidized by molecular oxygen in aqueous solution in the presence of species such

as chloride, which complex Hg(II); high chloride activities in Carson River and Lahontan

Reservoir sediments are a product of the prevailing arid climate conditions and

subsequent evapo-concentration processes (Wayne et al., 1996). This mechanism

facilitates mercury oxidation to occur in sediments that remain moist or in suspension,

including channel bed sediments and bank sediments under certain moisture conditions.









Other oxidizing agents within the sediment matrix and in the water, however, are likely at

work as well. The presence of sulfur within the ores of the Comstock era results in the

precipitation of some of the oxidized divalent mercury as relatively insoluble HgS in

highly-reducing environments, such as within the reservoir or buried portions of the soil

profile (Lechler et al. 1997). This mechanism serves to limit methylation of all oxidized

mercury, although very limited methylation of HgS may occur. Note that the prevalent

mercury species within the system remains the elemental form. The key item to take from

the above discussion, though, is the presence of inorganic, divalent mercury (Hg(II) or

Hg2+) within the near-channel and in-channel sediments of the Carson River system.

Chemistry of Mercury in Soil

Divalent mercury within soil is subject to many chemical and microbial reactions

(USEPA, 1997). For typical ranges of pH, temperature, and humic content in soils,

conditions generally favor the formation of inorganic Hg(II) compounds such as HgCl2,

Hg(OH)2, and inorganic Hg(II) compounds completed with organic anions. These

inorganic Hg(II) compounds possess high water solubility, but this theoretical mobility is

heavily limited by the tendency of these compounds to form complexes with soil organic

matter, primarily humic and fulvic acids, because of the affinity of Hg(II) and its

inorganic compounds for sulfur-containing functional groups. Because of the high

affinity of mercury species for soil, soil acts as a large reservoir for mercury introduced to

the environment and a leaching source that will continue for many years after the

cessation of mercury loading. Partition coefficients (Kd = concentration of Hg in dry

sediment, soil, or suspended matter divided by the dissolved concentration in water) have

been calculated for the relative affinity of Hg(II) and MeHg for sediment over water.

Typical values are 10 to 100,000 milliliters (mL) per gram(g) of soil, 100,000 mL/g of









sediment, and 100,000+ mL/g for suspended matter, indicative of a strong preference

Hg(II) and MeHg to remain in these phases (USEPA, 1997). In aquatic systems with

significant mercury loadings, it is possible for a significant amount of Hg(II) to partition

into the water column, especially in those instances where the concentration of suspended

material in the water column is high, as in the case of a flood event. Alluvial sediments

within the Carson River system are generally mineral-rich soils with low organic content.

In the absence of large amounts of organic material, complexation with these compounds

is limited; studies have suggested that enhanced mercury mobility within the Carson

River and Lahontan Reservoir system may also result from locally high chloride activities

(Wayne et al., 1996). Divalent mercury compounds and complexes can undergo reduction

to elemental mercury, processes mediated by both humic substances and light (USEPA,

1997). These processes will bring about diffusion of the gas and re-emission to the

atmosphere. Microbial activity on certain Hg(II) substances can result in MeHg

formation, which typically makes up between 1 and 3 percent of mercury found in soil

(USEPA, 1997). Most of the remaining mercury (97 to 99%) will be present as Hg(II)

species. Contaminated sediments may contain MeHg levels far greater than these typical

observations.

Mercury Transformation and Toxicity

Heavy metals, including mercury, possess a strong affinity for sorption onto

particulate matter given the Eh and pH conditions found in most aquatic environments

(Morel and Hering, 1993). Several contaminated river systems convey roughly 90% of

their total heavy metal load via sediment transport (Babiarz at al., 1998; Balogh et al.,

1997; Gill and Bruland, 1990; Hurley et al., 1995; Meade et al., 1995). The Carson River









is no exception in that more than 95% of total mercury can be removed from water

column samples by filtration (Bonzongo et al., 1996a).

Several forms of mercury are present within the sediments and waters of the Carson

River system. Due to its high volatility, only very low concentrations of elemental

mercury are detected in the surface waters of the Carson River (Bonzongo et al., 1996a).

Inorganic mercury and methylmercury in sediments along the Carson River are

transported into the water column through diffusion and erosion processes (Figure 2-2).

Mercury bound to particulate matter is generally considered unavailable for any chemical

or biological reaction and will be susceptible to sedimentation processes. The presence of

a state of sorption equilibrium (partitioning between sorbed and dissolved phases) does,

however, ensure the presence of a small quantity of free ionic mercury (Hg2+) within the

water column and pore water of saturated bed sediments and moist bank sediments. Free

ionic mercury is available for transformation. Within bottom sediments, a positive but

weak correlation between the ratio of total MeHg to total Hg and organic matter has been

observed (Hoffman and Thomas, 2000).

Sulfate-reducing bacteria (SRB) methylate free ionic mercury into the organic

mercury species methylmercury (CH3Hg) and dimethylmercury ((CH3)2Hg) (Gilmour

and Henry, 1991; Gilmour et al., 1992; Matilainen, 1995; Bonzongo et al., 1996a). The

production of MeHg by SRB is primarily controlled by SO42- concentrations required for

SRB activity. Optimal concentrations for the biotransformation of inorganic mercury by

SRB are estimated in the 200 to 500 ptM/L range; above this range, methylation is

inhibited by high production of sulfide, while lower concentrations inhibit methylation

because of sulfate availability (Bonzongo et al., 1996a). Evidence is available to support




























Figure 2-2. Processes that transport mercury into the water column along the Carson
River. (Reprinted with permission from Carroll, R.W.H., 1999. Simulation of
Mercury Transport and Fate in the Carson River, Nevada. Masters Thesis.
University ofNevada-Reno, Reno, Nevada. Figure 1.2, pp. 11)

abiotic contributions to the methylation and demethylation processes as well, but the

exact mechanisms remain poorly understood (Chen et al., 1996). While all mercury

species possess intrinsic toxic properties, it is only the organic mercury species that are

lipophilic and can bioaccumulate through the food chain (Fitzgerald and Clarkson, 1991;

Zillioux et al., 1993). Within freshwater systems, methylmercury accounts for the

majority, if not all, of the organic mercury species; the lack of dimethylmercury

accumulation in freshwater systems may be related to both its high volatility and

photochemical destruction (Fitzgerald and Mason, 1997). Methylmercury is therefore the

principal species of concern for both ecosystems and human health.

Sulfate-reducing bacteria are widespread in anaerobic bottom sediments but may

also reside within the anoxic hypolimnion of large lakes (Gilmour and Henry, 1991;

Matilainen, 1995; Watras et al., 1995b). Fluvial systems contain mostly oxic waters;

within the context of this study, only the anoxic channel bottom sediments and the moist


Mill Tailings

Bank Erosion

\ \Dghflows



Old Flood Plain Old Flood Plain


Diffusion









or saturated bank sediments (potentially anoxic environments) will be considered as sites

for mercury methylation.

The reader should further note that methylation and demethylation processes occur

simultaneously. Methylation refers to the biologically- or abiotically-mediated

transformation of free ionic (inorganic) Hg(II) to MeHg; demethylation is the conversion

of MeHg back to inorganic Hg. While demethylation reactions are kinetically

unfavorable, due in part to the presence of partially covalent metal-to-carbon bonds, the

unfavorable kinetics can be overcome through microbial enzymatic or photochemical

mechanisms (Amyot, 1998). While both processes must be modeled, interest is centered

on the net rate of MeHg production, as this is the determining factor for methylmercury

concentrations in channel bottom and bank sediments and (through subsequent diffusion

and erosion) the water column, the latter of which is available for introduction to the food

chain. The balance between methylation and demethylation processes may be critical to

maintaining low levels of MeHg in some surface waters. Processes which reduce the

availability of inorganic Hg(II) for methylation can further limit the potential for

biological uptake. Examples of these processes include the following: sedimentation of

inorganic Hg(II) compounds resulting in the precipitation of insoluble HgS as the

sediments become increasingly anaerobic; photochemical or microbial reduction of

Hg(II) to Hg(0), which is removed by volatilization, within aerobic surface waters; and

microbial or abiotic reduction of Hg(II) within anaerobic sediments (Amyot, 1998).

Methylmercury in the Water Column

The result of the chemical transformation of free ionic Hg2 to MeHg is the

introduction of methylmercury into the water column by diffusion and erosion processes.

Large flow events, when the Carson River erodes significant quantities of contaminated









bank material, are believed to be the principal pathway by which mercury enters into the

water column (and ultimately the Lahontan Reservoir and Wildlife Refuge areas).

Because mercury is adsorbed to the sediment, the increased erosion associated with high

flow events and the higher quantity of mercury present within the floodplain (compared

to the channel bottom) results in more mercury getting into the water column. Particulate

Hg is typically greater than 50% of total Hg and increased downstream; by contrast, the

dissolved fraction of MeHg constituted a larger fraction of total MeHg than

particle-associated MeHg (Bonzongo et al., 1996b). Flood events also contribute greatly

to the continued dispersal of mercury contaminated sediments throughout the river

system. Concentrations of both total Hg and MeHg in water are strongly influenced by

flow regimes and well correlated with concentrations of suspended solids, as result from

erosion (Bonzongo et al., 1996a; Bonzongo et al., 1996b).

Observations of MeHg and total Hg in the water column and bank sediments

indicate that the MeHg and total Hg concentrations measured in surface waters were

discharged in large part from Hg-contaminated bank materials, which are not in

permanent contact with the water body (Bonzongo et al., 1996a). Contaminants entering

the water column that are associated with silt and clay are transported to the reservoir,

while mercury associated with amalgam grains and other sand-sized sediment are

incorporated into the materials found beneath the channel floor (Miller et al., 1998).

Mercury also enters the water column by diffusion from the channel bed and contacted

bank sediments (Carroll, 1999; Carroll et al., 2000; Wayne et al., 1996). This source term

is thought to be particularly important for the introduction of MeHg to the water column

when the discharge is low and erosion is minimal. For water samples, positive but weak









correlations between total MeHg concentration and water temperature, pH, and total

organic carbon have been observed; in addition to a positive, moderate correlation

between total MeHg and total Hg concentrations. A strong positive correlation between

total MeHg in water and total Hg in sediment has also been observed (Hoffman and

Thomas, 2000).

The distribution and occurrence of Hg species in aquatic environments are probably

regulated by key limnological and chemical parameters. While significant concentrations

of methylmercury (MeHg) exist within the surface waters and biota of the Carson River,

the ratio of total MeHg to total Hg is much smaller than other contaminated sites (Bloom

and Effler, 1990; Bonzongo et al., 1996a; Gill and Bruland, 1990). The ratio of total

MeHg to total Hg decreases downstream, reaching values of less than 1% below the mine

tailing sites. This suggests both a high percentage of inorganic mercury input from point

sources and a low net product of methylation-demethylation within the system

(Bonzongo et al., 1996b). There is a strong indication that net methylation rates increase

with decreasing pH in both the water column and at the sediment-water interface (Watras

et al., 1995b; Xun et al., 1987). The alkaline nature of the Carson River (pH 7.5 to 9.0)

may limit the MeHg production. Naturally occurring Group VI oxyanions (W, Mo, and

Se) also inhibit the methylation process by SRB, and the extremely high levels of

inorganic mercury found in the Carson system during winter and spring flood events may

actually be toxic for SRB as well (Bonzongo et al., 1996a; Chen et al., 1996; Chen et al.,

1997).

The Carson River has a very low organic content (<2%) (Miller et al., 1998; Wayne

et al., 1996), and no correlation exists between DOC and total mercury or methylmercury









concentrations (Bonzongo et al., 1996a). Low levels of DOC limit complexation of

dissolved MeHg to dissolved organic ligands, and lead to the assumption that a large

fraction of dissolved MeHg is available for conversion processes and uptake by aquatic

organisms (Bonzongo et al., 1996b). Increased temperatures result in increased

methylation rates along the Carson River (Carroll, 1999; Carroll et al., 2000); this direct

relationship is strongly supported with several studies of other systems (Callister and

Winfrey, 1986; Matilainen, 1995; Stordal and Gill, 1995; Watras et al., 1995a).

Methylmercury within the water column is available for introduction to the aquatic

food chain. MeHg is introduced to the water column through bank erosion and diffusion

from contacted channel bank and bed sediments. Analysis of water samples from the

Carson River indicates that only a very small fraction of mercury found in the water is

available for conversion processes, including methylation, demethylation, and Hg(0) (or

Hgo) formation (Bonzongo et al., 1996b).

Total Hg and total dissolved mercury concentrations in the contaminated part of the

Carson River are 5 to 100 times higher than the following recently studied systems

polluted with mercury: Davis Creek, California; Clear lake, California; Lake San

Antonio, California; Pyramid Lake, Nevada; Onondage Lake, New York; and Clay Lake,

Ontario (Bonzongo, 1996b). Mercury concentrations in the Carson River are likely

among the highest in any aquatic systems, even those with known histories of Hg

contamination. However, the reactive fraction is observed to be lower within the Carson

system. This data is used to justify the exclusion of methylation-demethylation processes

within the surface waters of the Carson River model, limiting these reaction sites only to

bed sediments and saturated or moist bank sediments.









Methylation and demethylation kinetics are variable spatially within the Carson

River system and vary with the time of year. Preliminary findings from the USGS study

on methylmercury formation and degradation in sediments of the Carson River system

(CRS) include the following (Marvin-DiPasquale and Oremland, 1999):

* High rates of MeHg production in upstream river sediments and low rates in wetland
sediments;
* Low rates of MeHg degradation in the upstream river sediments and high rates in the
wetlands;
* Gross MeHg degradation strongly impacts net MeHg production;
* Microbial sulfate reduction and resulting reduced-S concentrations play an important
role in mediating Hg-transformations;
* Oxidative demethylation is the dominant degradation pathway in the CRS
* Agricultural drains are a potentially important source of mercury to the larger wetland
areas;
* Distinct spatial trends in microbial Hg-dynamics reflect the immediate environmental
conditions with respect to organic matter, sediment redox, pore-water anions, and
overall anaerobic microbial activity.

Note that most of these observations were discussed at greater length within earlier

portions of this document.

Mercury in the Aquatic Food Chain

The quantity of mercury transformed to methylmercury (methylated) can vary

greatly between water bodies because of several factors; some of these factors, other than

mercury loading rate, include the following: environmental setting (climate, geology,

land use, land cover), water chemistry, and wetland density, with the latter being the most

important determinant of methylmercury production at the basin scale (USEPA, 2000).

Methylation is a key step in the entrance of mercury into the food chain. This

methylation process occurs principally through biological processes in the aquatic

sediment or water column, but also perhaps through abiotic processes involving humic









and fulvic acids in solution. Demethylation reactions, volatilization of elemental mercury

and dimethylmercury, and bonding to particulate matter and settling prevent all of the

mercury loaded into a system from being bioavailable in the aquatic environment

(USEPA, 2000).

Microbial uptake of mercury is the first step for both possible methylation

processes and bioaccumulation considerations. Methylation by sulfate-reducing bacteria

or entry into the aquatic food chain via phytoplankton or bacteria first mandates the

transport of mercury across the lipid membrane that surrounds unicellular organisms

(Amyot, 1998). The microbial uptake of inorganic Hg(II) and methylmercury are affected

by elements of water chemistry. High chloride concentrations and low pH conditions that

favor the uncharged chloride complexes HgC12 and CH3HgC1, mercury complexes which

possess the highest lipid solubility, have been shown in field data to promote the

accumulation of inorganic Hg and MeHg at the microbial level by phytoplankton and

bacteria. Other non-polar mercury species, including dimethylmercury and elemental

mercury, also diffuse rapidly through cell membranes, but they are not bioaccumulated.

Bioaccumulation refers to organisms taking up contaminants more rapidly than their

bodies can eliminate them, allowing the amount in the body to increase over time.

In order for high concentrations to occur in fish, mercury must be efficiently taken

up by microorganisms at the bottom of the food chain, retained by these organisms, and

passed on to their predators (Amyot, 1998). Although trace metals are often accumulated

at the bottom of the food chain, concentrations in biomass do not often increase up the

food chain (a process called biomagnification). Biomagnification is the incremental

increase in contaminant concentration at each level of the food chain; this phenomenon is









the result of higher concentration of contaminant in the food sources for organisms higher

on the food chain (USGS, 1995). Methylmercury is very bioavailable and accumulates in

fish through the aquatic food web, with nearly 100 percent of mercury found in fish

muscle tissue in the methylated state (USEPA, 2000). Elemental mercury and

dimethylmercury are not bioaccumulated at the microbe level because they are

non-reactive and tend to diffuse out just as easily as they diffused into the cell membrane.

Further explanations are needed as to why MeHg bioaccumulates and Hg(II) does

not, despite both HgC12 and CH3HgCl diffusing into cells at similar rates, reacting with

components of the cell, and being efficiently retained (Amyot, 1998). Research was

conducted to explain this phenomenon on copepods feeding on marine diatoms. What

was shown is that Hg(II) is primarily bound to particulate cellular material of the diatom,

while MeHg is associated with the soluble portion within the diatom cell. Because the

particulate material is excreted by the copepod rather than absorbed, only the MeHg

associated with the soluble fraction is efficiently incorporated into the copepod. Other

field data has shown this difference in transfer efficiency as applicable to other

microorganism predator relationships, although perhaps counter-intuitive based on the

relative abundance of Hg(II) compared to MeHg found in water bodies.

This process continues up the food chain through zooplankton, and the lipid

solubility of MeHg allows it to be retained in fat tissues of these organisms. Further

explanation is, however, needed since once the trophic level of fish is reached, MeHg

burden in muscle tissue is of greater significance that in fat tissue (Amyot, 1998). In fish,

MeHg absorption occurs principally along the intestinal wall, whereas inorganic Hg(II) is

adsorbed at the microvilli interface (located on immune system lymphocytes that are









carried within the bloodstream); this results in a very low uptake rate of inorganic

mercury in comparison with methylmercury. This cycle results in increases in the

proportion of MeHg to total Hg from about 10% in the water column, to 15% in

phytoplankton, to 30% in zooplankton, and to 95% in fish. In fact, bioconcentration

factors (BCF = concentration of MeHg in biota divided by concentration of MeHg in

water column) are on the order of 105 to 106 (USEPA, 2000). The food chain structure

therefore plays a key role in mercury biomagnification, because the accumulation of

MeHg results mainly from the consumption of MeHg-containing food rather than uptake

from the water itself. As the number of trophic levels within an aquatic system is

increased, so too is the mercury concentration in the top predators of the food chain

(Amyot, 1998). Large saltwater fish with long life cycles, most notably sharks, tuna, and

swordfish, are likely to accumulate the highest amounts of MeHg (USDHHS, 1997).

Higher methylmercury contents are also typically found in fish species in more acidic

lakes (lower pH) (USEPA, 2000). Studies have shown that DOC concentration and pH

levels have a strong effect on mercury fate, with increasing acidity (decreasing pH) and

increasing DOC content resulting in higher body burdens in fish, perhaps by enhancing

the mobility of mercury in the environment and thus the likelihood of entering the food

chain (USGS, 1995).

Even though mercury loadings to some remote areas may be very low, the strong

absorption by microbial bacteria and plankton found in floating material make the

possibility of biomagnification increasingly likely if other conditions are suitable (USGS,

1995). Although mercury is a globally dispersed pollutant, problems do not arise

everywhere. With the exception of grossly polluted areas, mercury is generally only a









problem where the rate of natural formation of methylmercury from inorganic mercury is

greater than the reverse reaction (USGS, 1995). Specific environments known to favor

the production of methylmercury include certain types of wetlands, dilute low-pH lakes

in the northeast and north central United States, parts of the Florida Everglades, newly

flooded reservoirs, and coastal wetlands, particularly along the Gulf of Mexico, Atlantic

Ocean, and San Francisco Bay (USGS, 1995). Because methylmercury is the mercury

species of concern with regard to bioaccumulation and biomagnification within the food

chain, quantification of MeHg contamination within the Carson River is of great

importance.

Carson River Superfund Site

In August 1990, the stretch of the Carson River through Dayton Valley and

extending downstream to Lahontan Reservoir, to include Carson Lake, Stillwater

National Wildlife Refuge (NWR), and Indian Lakes in Lahontan Valley, along with

adjoining tributaries stretching nearly to Virginia City itself, was declared a Superfund

site by the U.S. Environmental Protection Agency (USEPA). The remediation strategy

for these sites remains unresolved. Lahontan Valley's wetland system encompasses a

large portion of the lower Carson River Basin and includes the Stillwater NWR, the

Fallon NWR, and Stillwater Wildlife Management Area. This area serves as a key

production, migration, and wintering area for up to one million waterfowl, shorebirds,

and raptors (NDWP, 1996). Mercury concentrations in soil samples are found to be some

200 times greater below the mill sites than above them, with the highest concentrations

found just upstream from Lahontan Reservoir (the principal area for sediment

deposition). Fish within the Carson River system have been found to have total mercury

levels four times greater than the Federal Action Limit for consumption of 1 [tg/g (Miller









et al., 1996; Carroll, 1999). Mercury levels in fish (up to 9.5 [g/g ) and waterfowl (>3.6

[tg/g) caught in and near the SWMA (Stillwater Wildlife Management Area) have also

been found to be well above the FDA limits for human consumption. The recreational

value of Lahontan Reservoir, which includes commercial and recreational fishing, and

the SWMA are severely impacted by the continued input of mercury. Abnormally high

Hg concentrations in sediments and fauna from the SWMA support a scenario of Hg

mobilization, transport, and methylation within the Carson-Lahontan system (Wayne et

al., 1996).

Health Effects of Mercury Contamination

Airborne and waterborne mercury may undergo several biological transformations

that can yield methylmercury. In this form, mercury efficiently accumulates in the aquatic

food web. Nearly all of the mercury that accumulates in fish in rivers, lakes, and

reservoirs is present in its methylated form. The methylmercury concentrations found in

fish specimens may be several orders of magnitude greater than concentrations in the

water column, the direct result of the bioaccumulation and biomagnification processes.

Inorganic mercury tends to not bioaccumulate (USEPA, 2000). The efficient

accumulation of methylmercury up the aquatic food web poses risks to both humans and

ecosystems. Other forms of mercury can also impact human health.

Toxic effects associated with methylmercury include permanent damage to the

brain and kidneys; and damage to the nervous system (USEPA, 2000). Mercury effects

many different brain functions and symptoms may occur which include the following:

personality changes (irritability, shyness, nervousness), tremors, changes in vision or

hearing, loss of sensation, and memory loss (United States Department of Health and









Human Services (USDHHS), 1997). Consumption of large amounts of

methylmercury-containing fish or seed grains treated with methylmercury or other

organic mercury compounds may cause permanent brain and kidney damage, as well as

damage to the brain of a developing fetus. Methylmercury exposure is most dangerous to

young children because a greater amount of the chemical passes into the brain and

interferes with brain development. Adding to the problem is that since methylmercury is

concentrated in fish muscle tissue (unlike organic contaminants like PCBs and dioxins,

which concentrate in the skin and fat), mercury cannot be filleted or cooked out of

consumed fish (USGS, 1995).














CHAPTER 3
PREVIOUS MODEL STRUCTURE

Mercury fate and transport within the Carson River are simulated using three

computer models: RIVMOD (RIVer MODule), WASPS (Water Quality Analysis

Simulation Program), and MERC4 (MERCury). These models were originally chosen

and linked by Heim and Warwick (1997). Currently, RIVMOD is a subroutine within the

WASPS code and is executed during each model time step. Sediment transport is

simulated within the WASP5/MERC4 system of models and not within the RIVMOD

code. The resulting model framework is referred to as the Integrated Fluvial Mercury

Transport and Fate Model, or IFMTFM.

Hydrodynamic Modeling Using RIVMOD

RIVMOD (Hosseinipour and Martin, 1990) is a USEPA one-dimensional (ID)

hydrodynamic and sediment transport routine that simultaneously solves the standard

fluid equations of continuity and momentum throughout the system being simulated,

utilizing the Gauss-Seidel iterative scheme. Finite difference equations are solved by the

Newton-Raphson method to determine fluid velocity and depth given unsteady flow

conditions. RIVMOD's input requirements include bed elevation, cross-sectional

geometry, roughness coefficients, and initial conditions for water depth and discharge.

Discharge is specified at the upstream and downstream boundary. The model domain

includes the Carson River from the USGS Carson City Gage (CCG) through Lahontan

Reservoir. The 115-km, or 71.4 mile, reach is subdivided into 273 segments. The location

of each model segment is specified by model stations located at both the upstream and









downstream ends of the segment, resulting in a total of 274 stations. Station 1 is located

at CCG and Station 274 at the base of Lahontan Dam. Station spacing is generally 0.50

km throughout the modeled reach, although station spacing is reduced to 0.25 km in the

transition region between the river and reservoir for stability reasons.

Discharge is specified at the upstream boundary based on CCG data. At the

downstream boundary, discharge from Lahontan Reservoir is specified with data from the

USGS gaging station below the dam (LDM #10312150). Inflow from the Truckee Canal

into Lahontan Reservoir is specified with data from the USGS gaging station on the

Truckee Canal near Hazen, Nevada (TRC #10351400). Data available from the USGS

gage at Fort Churchill (FCH #1031200) is used to define a time-varying lateral outflow at

this location (IFMTFM station 128); the flow difference between what is predicted and

what is measured at FCH is removed, thereby accounting for irrigation diversions and

loss/gain due to groundwater interaction between the upstream and downstream

boundaries. Here it is important to note that no diversion ditches are in operation

downstream of the FCH gaging station.

Two major modifications have been made to RIVMOD. The original RIVMOD

code was capable of modeling sediment transport, but geometry was limited to a simple

rectangular channel or a first- or second-order polynomial. The first modification

implemented a more complex cross-sectional geometry. Figure 3-1 presents the channel

representation incorporated into the modified version of RIVMOD (Warwick and Heim,

1995). Note the cross-section is divided into two basic components: the main channel

(parameter BW2) and floodplain (depth greater than D2). The inner channel and

low-medium flow transition slope portions of the cross-section represent the channel bed.











The portion of D2 above the low-medium flow transition slope represents what will be


referred to as the channel bank; the modeled portion of the Carson River is characterized


by steep, nearly-vertical channel banks. The RIVMOD input file includes values for each


parameter at each model station, the combined result of averaged field measurements and


approximate values as taken from USGS topographic maps. The modified version of the


code is able to consider changing flow and cross-sectional area between adjacent stations.


The second modification applied a divided channel approach to the momentum equation,


thus allowing for a separate floodplain roughness coefficient and the computation of


overbank depths and velocities (Carroll, 1999). A smaller modification was also made to


allow for the inclusion of evaporative losses, which are potentially significant in the


semi-arid region of the Carson River system.


BW4

BW3

BW2


DI = Low flow inner channel depth (0.20 m)
D2 Main channel depth (4.0 m)
BWI = Low flow inner channel width (3.0 m)
BW2 = Main channel width (45 m)
BW3 = Floodplain transition zone width (61 m)
BW4 = Inner floodplain width (914 m)


SLI = Low medium flow transition slope (0.10)
SL2 = Medium high flow transition slope (0 07)
SL3 inner floodplain slope (0.07)
SL4 = Outer floodpl.un lope IN) I


Figure 3-1. Modified channel geometry


i-









The WASPS and MERC4 Programs

WASPS is the USEPA Water Quality Analysis Simulation Program-5 that was

developed to simulate the transport and transformation of various water body constituents

(Ambrose et al., 1991). Mass balance equations account for all material entering and

leaving model segments through direct and diffuse loading, advective and dispersive

transport, and any physical or chemical transformation. Input parameters required by

WASPS include simulation and output control, model segmentation (same as RIVMOD),

advective and dispersive transport variables, boundary concentrations, point and diffuse

source waste loads, and initial conditions.

Several modifications have been made to the WASPS numeric code. The RIVMOD

model is currently a subroutine within the WASPS program. This linkage allows for

sediment transport information, as related to hydrodynamic characteristics, to be

incorporated into the WASPS mass flux calculations (Heim and Warwick, 1997). The

current WASPS model simulates the transport of three distinct solid types: washload

(diameter < 0.063 mm), coarse suspended sediment, (diameter > 0.063 mm), and

bedload. Washload and CSS (coarse suspended sediment) are carried in suspension,

while bedload is defined as coarse material that travels by rolling, skipping, and/or

sliding along the riverbed. Washload and CSS boundary conditions are modeled by

making use of measured rating curves at CCG; by contrast, bedload is modeled using van

Rijn's transport theory (van Rijn, 1984a; van Rijn, 1984b; Carroll, 1999). USGS data

indicate that CSS is held constant between CCG and FCH. While exchange is modeled

between water column and bed segments, there is no net increase in CSS. However, data

also show that on average 70% of the bank material is considered coarse (Heim and

Warwick, 1997). Thus, for the purpose of modeling the system, it is assumed that eroded









coarse bank material settles onto the bed segment and is transported as bedload. The

exchange between bedload and CSS is modeled by satisfying the water column CSS

concentrations as defined by the measured rating curves. Equations governing the erosion

of washload were originally developed by Carroll (1999). Modifications to these erosion

equations and additional relationships for modeling overbank deposition during extreme

flood conditions were later introduced by Carroll et al. (in press). It is assumed that the

river is in vertical equilibrium with erosion and/or deposition occurring only laterally

across the floodplain and not along the channel bed. Channel bank and floodplain are

subject to dynamic erosion processes within the IFMTFM, while channel bed is assumed

to remain constant. It is further noted that channel geometry is not updated within the

RIVMOD code following modeled sediment erosion.

MERC4 is a subroutine contained within WASPS (Martin, 1992). It was developed

to compute mercury speciation and kinetic transformation. MERC4 is capable of

simulating up to four mercury species: elemental Hg(0), Hg(2+), inert Hg, and

Methylmercury (MeHg). The MERC4 model is also capable of simulating up to three

solids types: washload, CSS, and bedload in this case. For this study, only inorganic

mercury and methylmercury are modeled. Note that only the soluble forms of these

mercury species are available for chemical transformation. The simulation of solids is

critical because of the strong tendency of mercury to sorb onto solids. The mercury

speciation and kinetics algorithms are based on the Mercury Cycling Model (MCM)

developed by Tetra Tech, Inc. (Hudson et al., 1994) for the Electric Power Research

Institute. The MERC4 model resulted in the coupling of chemical speciation and kinetics,









based on the MCM model, with the transport capabilities of the WASP4 modeling

system.

Within the MERC4/WASP5 program, mass balance equations are written for each

form of mercury and solids being modeled. The initial total concentrations for each of

these state variables are specified to the model. New total concentrations are computed

based on transport, transformations, and reaction kinetics, which vary depending on the

species present (Martin, 1992). In addition to the state variables, other chemicals which

affect mercury speciation are input to the model but not predicted, including the

following: pH and concentrations of free sulfide, sulfate, chloride, dissolved organic

carbon, and dissolved oxygen. The chemical species considered by MERC4 include

dissolved, sorbed, and precipitated forms.

Measured rating curves at CCG for inorganic mercury and methylmercury in the

water column were written into WASPS as an upstream boundary condition (Carroll,

1999). Reaction constants for sorption (Bonzongo et al., 1996a); and methylation and

demethylation reactions (Carroll, 1999; Carroll et al., 2000), which are input into the

model, are based on field data collected along the Carson River system. Initial conditions

for mercury bed concentrations are described by fitting a three-point moving average

through data collected by Miller and Lechler (1998). Initial concentrations for mercury in

the banks were based on calibrated parameters and were not dynamic in nature. The

previous version of WASP5/MERC4 does not simulate diffusion other than from the bed

sediments. These shortcomings of the model in terms of simulating field conditions will

be addressed later in this chapter.









Stages of Model Development

The modified and linked RIVMOD/WASP5/MERC4 model, as described above,

makes up the framework of the IFMTFM. Model development has occurred in three

distinct, but ongoing phases (Heim and Warwick, 1997). The first phase of the project

applied RIVMOD to the Carson River and Lahontan Reservoir system (CRLR) to

simulate a variety of river flows and reservoir stages (Warwick and Heim, 1995). The

second phase incorporated sediment transport and characterization of the association of

mercury with sediment transport. The third phase seeks to incorporate the results of the

first two phases with MERC4 to determine areas of potential mercury methylation, define

important transport mechanisms, isolate sources or times of increased mercury

contamination, and allow application of the model to different hydrologic scenarios. A

working model may then be used to test the outcome of mitigation strategies, such as the

removal of mercury sources. This effort may ultimately lead to an effective way to reduce

mercury concentrations in the water column, and ultimately the food chain.

While much of this research was previously completed (Heim and Warwick, 1997;

Carroll, 1999; Carroll et al., 2000; Carroll et al., in press), this thesis improves the

capabilities of the IFMTFM through the incorporation of bank elements. The purpose

behind the development of the bank elements to be integrated into the current IFMTFM

code is to develop a means to simulate the spatial and temporal variability of

methylmercury within the channel banks of the Carson River system by better simulating

the actual physical and biological processes which are occurring in the field.

Methylmercury concentrations in the bank are not spatially and temporally constant but

are instead a function of moisture history. A modified Carson River model is developed

to include the relation between flow history and moisture levels observed within the






37


banks. Zones of varying moisture impact methylation and demethylation kinetics, and

therefore the temporal concentrations of methylmercury available to the water column.














CHAPTER 4
CHANNEL BANK ELEMENTS: LINKING SURFACE WATER AND
GROUNDWATER

Modeling the Carson River

The USEPA designated the Carson River, Nevada as part of a Superfund site in

1991. Between 1860 and 1890, elemental mercury was used at over 70 mill sites to

recover gold and silver via mercury amalgamation, introducing 7000 tons of mercury to

the Carson watershed. Contaminated sediments line channel banks downstream of these

mill sites, with mercury concentrations two to three orders of magnitude above

background levels. Mercury methylation within sediments, accompanied by diffusion and

erosion, introduces methylmercury to the channel. Methylmercury concentrations in the

bank are not spatially and temporally constant but are instead a function of moisture

history. A modified Carson River model is developed to relate flow history to moisture

levels observed within the banks. Zones of varying moisture impact methylation and

demethylation kinetics, and therefore the temporal concentrations of methylmercury

available to the water column.

Methylmercury concentrations within the Carson River channel banks vary with

both spatial location and time. The temporal variability in methylmercury concentrations

was not previously considered by the IFMTFM. The previous modeling approach (prior

to this thesis) considered methylmercury bank concentrations as temporally constant and

inversely related to the square root of channel bottom slope (So) (Carroll, 1999):






39

[HgMeHg] 2
S^O. (Eq. 4-1)

A calibrated threshold flow of 1000 cfs was developed, below which methylmercury is

available to be eroded from the channel banks. This threshold flow was later modified to

a threshold depth of 1.00 meters. Diffusion of methylmercury was only allowed to occur

from the channel bottom of the main channel. No efforts are included within the previous

model to simulate groundwater hydrodynamics in response to flow history within the

river channel. It is this temporal variability in bank saturation zones that controls the

kinetics underlying the microbial production of methylmercury within the channel banks,

thereby influencing MeHg available for introduction to the water column. Research

concerning the relation between soil wetting and drying cycles; and methylation /

demethylation kinetics is ongoing at other academic institutions.

The end product of this thesis research is the introduction of channel banks into the

IFMTFM code as areas for active mercury transformation. Channel bank elements are

created within the IFMTFM, conceptually similar in functionality to bed elements within

the IFMTFM. Bank element size is dynamic and equal to the saturated portion of the

bank. Methylation and demethylation will occur within channel banks, resulting in

dynamic methylmercury concentrations. As in the previous IFMTFM version, channel

banks are eroded, but the methylmercury concentrations within the eroded material are no

longer temporally constant. The code was already able to predict flow magnitudes and

depths along the Carson River from the Carson City Gage through the Lahontan

Reservoir. The creation of bank elements, along with the accompanying coding

modifications, allows the model to predict the zone of saturation within the channel

banks. The functionality of these bank elements is determined based on analysis of the









following: river stage predictions as computed from a 3-dimensional (3D) hydrodynamic

model, water table data from three groundwater wells, and the time lag between river

stage oscillations and observed water table response.

Available Data

To obtain more precise channel geometry data, a GPS (global positioning system)

unit, the Leica GPS System 500 (Leica Geosystems, Atlanta, Georgia), was purchased

and used to collect detailed data for a 0.5 km reach of the lower Carson River, with this

reach encompassing the location three installed groundwater wells (Scout Camp field

site). Signals from at least four satellites are needed by a GPS receiver on the ground to

resolve measurement ambiguities and obtain the location (latitude, longitude, and

elevation) of a given observed point. Without getting into the complexities of the system,

observations from four satellites are sufficient to resolve ambiguities and obtain spatial

location on the surface of the earth. The repetition of this procedure allows for the

collection of detailed channel morphology data for this small reach of the Carson River.

Vertical resolution of data points, based on system specifications, is reported to be on the

order of a few centimeters.

Flow data (magnitude and depth) from the USGS gaging station at Fort Churchill,

water table data (March 2001 through present) from the three wells installed along a

channel transect, and GPS survey data are used to construct a detailed surface-water

model and to integrate channel and groundwater hydrodynamics. Specifically, the goal

was to develop simple relationships between flow history (specifically rising and falling

water depths) in the channel to moisture content values (specifically the height of

saturation) within the channel banks. The simple relationships discovered from modeling

efforts and data analysis were incorporated into the IFMTFM modeling of channel bank









elements. With the IFMTFM model then able to predict bank moisture content both

spatially (due to variable channel geometry) and temporally, microbial kinetics data was

included to allow the model to predict spatial gradients and temporal variability of

methylmercury concentrations within the banks of the Carson River. The inclusion of

bank elements allows the IFMTFM to more accurately simulate mercury concentrations

present in eroded bank material and methylmercury gradients for diffusion processes.

Calibration and verification of the model is used to compare the updated IFMTFM model

with data collected within the Carson River system. Although the bank element is

conceptually simple, a historical perspective with regard to modeling of surface water

and groundwater interaction is important to gain additional insight prior to a discussion

on data analysis.

Introduction to the Hyporheic Zone

Historically, much ecological and hydrological research on groundwater and rivers

has treated the two systems as distinct entities. River and groundwater systems are, in

fact, characterized by marked differences (Boulton et al., 1998). Rivers typically have

currents generating turbulence; short water residence time; variable discharge and

physicochemical conditions; dynamic channel morphology; unidirectional transport of

nutrients, sediments, and biota; and are open to sunlight. By contrast, alluvial

groundwater environments tend to offer increased stability, longer water residence times,

laminar flow regimes, little change in sediment bed structure, and permanent darkness.

Despite these differences, the interface between surface streams and alluvial groundwater

represents a region of active interchange of water, nutrients, biota, and other materials.

This intervening zone is termed the hyporheic zone. Exchange occurs because of

variations in discharge, bed topography, and porosity (Boulton et al., 1998). Upwelling









groundwater supplies nutrients to stream organisms, while downwelling stream water

provides dissolved oxygen and organic matter to microbes and invertebrates in the

hyporheic zone (where photoautotrophic production is not possible). Dynamic gradients

exist at all scales (due largely to sediment heterogeneity), and the boundaries of the

hyporheic zone vary in both time and space. Flow occurs in three dimensions:

longitudinally down the channel, vertically between the channel and underlying

sediments, and laterally to or from the riparian zone (Jones and Holmes, 1996).

Subsurface regions are classified as hyporheic or groundwater zones based on hydrologic

exchange with surface waters; one definition proposes the division between the hyporheic

zone and groundwater to be where less than 10% of the subsurface water originated from

the surface channel (Fuller and Harvey, 2000; Jones and Holmes, 1996).

The dynamics of the hyporheic zone are particularly important for mercury

speciation. As the extent of the hyporheic zone fluctuates with stream stage and the

associated changes in hydraulic gradients, so too do the anoxic zones where the

methylating population is able to thrive. Flow rates through sediments are much slower

than on the surface, thus constraining the import of dissolved oxygen. If oxygen demand

from aerobic metabolism and other chemical reactions exceeds the influx, anoxic regions

will develop and have two important effects on hyporheic zone processes (Jones and

Holmes, 1998). First, organic matter decomposition occurs via anaerobic pathways, such

as denitrification, sulfate-reduction, and methanogenesis. Secondly, the oxic-anoxic

interface is the site of redox/precipitation reactions, such as the oxidation of methane or

ferrous iron. Additionally, methylating SRB thrive in these anoxic environments. The

active exchange of water in the hyporheic zone with water in the stream enables









methylmercury introduction to the biota of the channel and offers an additional

introduction pathway to bank erosion. The contribution of hyporheic zone processes to

the stream ecosystem function is determined by the rate of chemical transformations in

the sediments and the proportion of stream discharge flowing through sediments (Jones

and Holmes, 1996). It should be further noted that flow of surface water in the hyporheic

zone increases the effective reactive site density per volume of surface water by

providing contact of dissolved metals with potential reaction or sorption sites, thus

potentially enhancing metal attenuation by transport into zones where removal processes

are more favorable (Fuller and Harvey, 2000).

Modeling Surface-Subsurface Interactions: A Historical Perspective

A number of different approaches were taken over the years in the effort to quantify

stream-aquifer hydraulics. Individual efforts to obtain analytical expressions to relate

stream and aquifer hydraulic heads are each based on the same starting point, the

continuity equation. The differences arise in the assumptions made and the consideration

of different boundary conditions. In the interest of completeness, a number of these

approaches are summarized within this section.

Groundwater levels in alluvial valley aquifers vary with stage changes in the

associated surface streams. A rise or fall of stream stage over a short time period imparts

a flow reversal within the aquifer, a result of the change in gradient between the stream

and the aquifer (Serrano and Workman, 1998). The distance of influence within the

aquifer is dependent on the aquifer transmissivity and porosity, the change in stream

stage, and the length of time that conditions for a flow reversal exist (Serrano and

Workman, 1998). The hydraulic gradient is also spatially variable; river stage typically

does not increase uniformly over a long reach because of variations in channel geometry









and bottom slope (Squillace, 1996). When water quality within the stream differs from

that within the aquifer, the spread of contaminants (e.g. methylmercury) is a problem

directly connected to the hydraulics of the stream-aquifer system.

A rise in river stage, as associated with downstream movement of a flood wave, is

accompanied by lateral stream losses to bank sediments. As the flood wave passes, the

volume of water held in bank storage is gradually released to sustain baseflows between

flood events. Channel banks and floodplains thus mitigate against extreme hydrologic

phenomena by storing substantial volumes of water which would otherwise increase

flood volumes, a process known as bank storage (Whiting and Pomeranets, 1997;

Hantush et al., 2001). While this interaction between surface streams and groundwater

has often been studied in the interest of flood control, the underlying equations remain

the same for gradually varied stage changes within the stream.

The discussion to follow includes mention of a number of solutions developed for

unconfined aquifers. Yeh (1970) and Marino (1975) provide numerical solutions by

accounting for streams as time-varying boundary conditions. Further numerical solutions

have been developed modeling the stream and aquifer as interacting dynamic systems,

including work by Pinder and Sauer (1971) and Zitta and Wiggert (1971).

One-dimensional open channel flow equations are coupled with the two-dimensional

(2D) ground water flow equation. Assumptions embedded within these solutions may

bring about questions as to their practicality. Despite being more restrictive, practical

analytical solutions have also been developed. Several authors have presented analytical

solutions to the linearized Boussinesq equation, subject to a fluctuating stream stage.

Cooper and Rorabaugh (1963) derived an analytical solution for the changes in









groundwater head when the river stage can be approximated as a damped sinusoid, as in

the case of a passing flood wave. A finite difference expression of the analytical solution

developed by Cooper and Rorabaugh was developed by Hornberger (1970). Hall and

Moench (1972) and Moench et al. (1974) use the convolution integral to expand

applicability to flood waves of arbitrary shape, as described by a step function. A more

general solution developed by Workman et al. (1997) includes the effects of transient

redistribution of water levels in the aquifer from the previous day. Govindaraju and

Koelliker (1994) assessed the accuracy of these analytical solutions compared with

numerical solutions of the nonlinear equation, finding as much as 10% error.

The analytical solutions do not consider the simultaneous interactions that are

inherent to a dynamically coupled stream-aquifer system; rather, their use is based on a

prior knowledge of the stream-stage fluctuations, and they cannot be used for routing

streamflows without iterative procedures (Hantush et al., 2001). One key source of

difficulty in the efficient numerical coupling of streamflow and aquifer equations is the

intrinsic difference in the time scales of water movement in the two systems (Perkins and

Koussis, 1996). Analytical solutions to simultaneously couple the governing equations of

stream flow and groundwater flow are developed by Morel-Seytoux (1975), Hunt (1990),

and Harada et al. (2000). Analytical solutions to the non-linear Boussinesq equation are

later presented by and Serrano and Workman (1998). Each of the above solutions was

developed starting from the following standard assumptions, with additional assumptions

included as needed (Sharp, 1977):

* The alluvial aquifer is homogeneous, isotropic, and infinite in extent.
* The alluvium's bottom boundary is horizontal and impermeable.
* The Dupuit-Forchheimer conditions are valid. These conditions are:
o In any vertical section the groundwater flow is horizontal;









o The velocity is uniform over the depth of flow;
o The slope (a) of the free surface (unconfined flow) is small enough that
tan (c ) sin ( ).
* All media are saturated.
* Water from storage is discharge instantaneously upon reduction in head.
* Fully-penetrating streams (i.e., the streams penetrate to the bottom of the alluvium, or
alternatively, the heads in the alluvium beneath the river bed are all times equal to the
river stage.
* A solely unconfined or solely confined groundwater system.

Some authors have sought to develop more general solutions by relaxing some of

these assumptions. Solutions derived through use of linearization techniques applied to

the nonlinear Boussinesq equation are naturally expected to contain some error, caused

by the nonlinear nature of the original equation (Gill, 1985). This being the case, one may

choose to use even simpler solutions, using suitably calibrated values of the aquifer

constants to obtain reasonable results. Gill (1985) deduced limiting solutions for finite

aquifers of long-thin and short-thick dimensions, rather than assuming aquifers of infinite

thickness. Hantush (1967) derived 2D, time-dependent equations for a stream-leaky

aquifer system subject to an instantaneous rise in stream water level, with a constant

water level thereafter. Dever and Cleary (1979) advanced Hantush's approach with

general solutions for any stream stage hydrograph that can be represented by a

continuous or piecewise continuous function of time. An analytical model of

stream-aquifer interaction that considers the effects from a small degree of aquifer

penetration and low-permeability sediments on the head response to an arbitrary stream

stage hydrograph was proposed by Zlotnik and Huang (1999).

The progression of research into the area of stream-aquifer interaction led to the

development of numerical models, which became required tools as assumptions were

relaxed and solutions became increasingly complex and realistic. Neuman and









Witherspoon (1971) developed an iterative, finite element method applicable to a wide

variety of problems involving complex boundaries and arbitrary degrees of heterogeneity

and anisotropy. This approach, which considers 2D flow within channel banks (and

thereby does not encompass the limitations of Dupuit assumptions), was used in the

development of a bank storage model (WaTab2D) by Whiting and Pomeranets (1997).

Two-dimensional flow within channel banks for unconfined aquifers, of either semi-

infinite or finite width and without a semipervious streambank, was further considered in

work by Moench and Barlow (2000). It was necessary to assume that water table

fluctuations were small relative to the saturated thickness and that vertical flow into or

out of the zone above the water table occurred instantaneously, therefore neglecting the

effects of hysteresis in the moisture distribution above the water table. In addition to

considering stream interaction with anisotropic, homogeneous unconfined aquifers,

Moench and Barlow (2000) also considered stream interaction with various homogeneous

confined and leaky aquifer types. The computer programs STLK1 and STWT1 were

developed for simulations of confined and leaky aquifers and confined and water-table

aquifers, respectively (Barlow et. al., 2000). It was noted that a water-table aquifer

(unconfined) may respond very much like a confined aquifer if the unsaturated zone is

thin and the specific yield is small, the result of the non-instantaneous release of water

from storage following a reduction in head. Freeze (1971) was one of the earliest

researchers to consider the integration of the saturated and unsaturated zones into a

transient model. One common failure with the use of non-integrated fully saturated flow

models is that slower water table responses are predicted than are actually observed

(Wise et al., 1994).









In the choice of a model for a given system, it is important to consider the

assumptions upon which the model is based. Assumptions made for some of these

solutions are likely to be questionable in some instances and can yield erroneous results.

These questionable assumptions include Dupuit-Forcheimer conditions, homogeneity,

full penetration of streams, and systems that are solely confined or unconfined (Sharp,

1977). Sharp used data collected from major alluvial valleys in the central United States

to assess the validity of these assumptions; his findings are listed below:

* The assumption of homogeneity is clearly invalid as demonstrated by the exponential
increase in hydraulic conductivity with depth.

* Isotropy assumptions might be justified in the deep alluvium; anisotropic conditions
may occasionally be present at shallower depths.

* Valley bottoms may be adequately represented by a horizontal surface; the extremely
high hydraulic conductivities of the deep alluvium also justify the assumption of
impermeable boundaries, at least for first approximations.

* It is clearly shown that Dupuit-Forcheimer assumptions are not valid in the major
alluvial systems considered. Flow is predominantly vertical in the vicinity of the river
bed; groundwater flow is also not uniform with depth. While the approach reduces the
problem's dimension by one, it does not allow accounting for the vertical component
of the seepage velocity or accurate prediction of a significantly sloped free surface
(Whiting and Pomeranets, 1997).

* The assumption of saturated flow is legitimate and typical conductivities allow for
very rapid (instantaneous) release of water from storage upon head reduction.

* The rivers of the major alluvial aquifers studied are not fully penetrating. Presumably,
most groundwater flow between the river and bank storage is through the river bed
and not the banks because:

o The river bed has greater surface area, and
o The river bed is closer to the zone of greater hydraulic conductivity.

Additionally, the banks and bed, in general, will have differing hydraulic
conductivity because of smaller sediment sizes and higher clay percent in the bank
than in the bed, further decreasing the contribution of the banks to the river-aquifer
interaction. This condition leads, in part, to the invalidity of the Dupuit-Forcheimer









assumptions and makes most ID models suspect. The simulation of bank storage
using semi-permeable stream banks is not directly applicable.

* Alluvial aquifers may be confined and unconfined and these conditions may vary
both spatially and temporally.

Based on the above discussion, it becomes apparent that many bank storage models

are based on assumptions that do not correspond to observed field conditions. As noted

by Sharp, this does not necessarily invalidate the models. Rather, the validity of a model

and its assumptions rests on two divergent sets of criteria: "The positivistic view states

that any assumption which leads to a useable predictive model is a valid assumption...the

assumptions are valid regardless of any correspondence to observed or theoretical

processes. The normative view states that assumptions should correspond to the actual

processes or the true relationships between variables. If not, the assumptions are invalid

because they yield or are faulty interpretations" (Sharp, 1977). The listed assumptions fall

into both classifications, with difficulty arising when positivistic assumptions are used to

infer the actual hydrogeologic processes or when data from a positivistic model is

extrapolated to a situation where conditions differ by any significant degree (Sharp,

1977). "The problems involved in modeling without simplifying assumptions are

self-evident. Hydrogeologic variability insures that models cannot precisely correspond

to real-world processes. Nevertheless, the hydrologist or geologist who is attempting to

describe or model and bank-storage situation must be aware of the limitations of his

assumptions. The degree to which invalid (in the normative sense) assumptions reduce

the accuracy or interpretation or prediction will depend on the degree of divergence from

the real world and cannot be predicted a priori" (Sharp, 1977).









Modeling Simplifications

Despite the probable presence of a regional groundwater flow perpendicular to the

Scout Camp groundwater well transect, limiting data requires the assumption that water

fluxes along the front and back boundaries of this cross-section are negligible in

comparison to fluxes along the cross-section (perpendicular to the stream). River stage

and water table levels are then able to be analyzed as a 2D model. It is essentially

assumed that changes in hydraulic head within the aquifer are due primarily to the

pressure head gradients associated with the changing river stage, rather than changes in

aquifer head levels further upstream. This phenomenon is shown schematically in Figure

4-1.


River
stage
^ W Tater mble?
allei i -6 ii grI3

AHO / --


S I Water table
before the change HX
H







Figure 4-1. Water table response to changing river stage

Using IFMTFM Modifications to Track Bank Saturation: Bank Element
Functionality

Knowledge in hand, the functionality of the IFMTFM model is next considered;

specifically, how the model previously operated is compared with the updated model









with incorporated bank elements. Sediment transport and mercury transport and kinetics

within WASP5/MERC4 are reviewed. The previous IFMTFM includes two types of

elements: water column elements and bed elements. The modeling of mercury cycling

within channel banks required the creation of a new type of element: the bank element.

The bank element is used to track moisture content levels within the channel banks. As

moisture contents change within bank elements, properties of these elements relating to

mercury modeling are updated. The ability to model moisture content variability within

the channel banks, utilizing the newly-developed bank elements, allows for a more

detailed analysis of microbial behavior concerning methylation and demethylation.

Mercury modeling within the previous IFMTFM was accomplished as follows

(Carroll, 1999). The Carson City Gage (CCG) measured rating curves for washload and

coarse suspended sediment, along with a calculated bedload rating curve, are written

directly in the WASPS code as upstream boundary conditions, or rather, input to the

model domain. Measured rating curves for inorganic mercury and methylmercury in the

water column are also directly written into the WASPS code as upstream boundary

conditions. Mercury concentrations within the channel banks were chosen to calibrate the

water column concentrations of inorganic mercury and methylmercury. Miller et al.

(1999) found a statistically significant correlation between mercury concentrations and

the percentage of fine sediment (washload, diameter < 0.063 mm), with the implication

being that the contaminated material is primarily associated with the fine sediment. This

fine sediment is deposited along the lower reaches of the Carson River where the stream

gradient is shallow and velocities are slow. The assumption was therefore made that









concentrations of inorganic mercury and methylmercury within the bank sediments are

inversely related to the slope of the channel:


[Hg, ]bank (Eq. 4-2)

H MeHg ank (Eq. 4-1)
0

So is the channel bottom slope and i1 and k2 are calibration parameters adjusted to match

water column mercury concentration data. Bank mercury concentrations (inorganic and

methylmercury) are not dynamically changed in the previous version of the model. This

IFMTFM version includes a threshold flow depth (based on segment geometry and

calibrated QT = 1000 cfs) above which methylmercury does not occur in the channel

banks because methylating microbial communities cannot survive.

Within the confines of the previous model, methylation and demethylation

(resulting in dynamic concentrations of MeHg and inorganic Hg(II)) are allowed to occur

in the channel bed sediments (confined by the main channel) and these kinetic reactions

are governed by the equations 4-3 and 4-4:

[ =P ftepK2 [C, ] (Eq. 4-3)
at
T-20
temp Q1010 (Eq. 4-4)

For mercury methylation, the reactant (Cr) is soluble Hg2 and the product (Cp) is soluble

methylmercury. In the case of demethylation, soluble methylmercury serves as the

reactant and soluble Hg2 serves as the product. K20 is the reaction rate (day1), ftemp is a

temperature multiplier to correct for temperatures deviating from 20C dimensionlesss), t

is time in days, and reactant and product concentrations are in units of milligrams per liter









(mg/L). The spatial variability of the constants K20 and Qio are summarized in Table 4-1

(Carroll, 1999; Carroll et al., 2000).

Table 4-1. Spatial variability of reaction constants within IFMTFM
Range Methylation Demethylation
(distance from CCG in km) K20 (day-') Q10 K20 (day-') Q10
0.00- 79.25 0.0041 2.03 0.4483 2.32
79.25 95.40 0.0011 2.03 1.3291 2.32
95.40- 115.00 0.0028 2.03 1.1753 2.32

Initial mercury concentrations in bed sediments were determined by fitting a

three-point moving average through data collected by Miller and Lechler (1998) (Carroll,

1999). Inorganic mercury and methylmercury were introduced to the water column

through the following mechanisms (Figure 2-2): erosion of channel bank (vertical portion

of channel at BW2 in Figure 3-1) sediments; and diffusion from bottom sediments within

the confines of the main channel (BW2 in Figure 3-1) (inner channel is assumed in

dynamic vertical equilibrium and therefore not subject to erosion). Only deposition, not

erosion, is considered to occur during flows which contact the floodplain (BW2 to BW3

and BW3 to BW4 in Figure 3-1). Mercury concentrations within the eroded sediments

remain constant over time, although concentrations are spatially variable based on the 1i

and k2 calibration parameters. Water column concentrations are based on measured

partitioning constants between dissolved and adsorbed phases. MeHg and inorganic

Hg(II) are the only species modeled, based on the high volatility of elemental mercury.

Methylation and demethylation only occur within the water-contacted bed sediments of

the main channel, based on the predicted free ionic MeHg and free ionic inorganic Hg(II)

concentrations. Free ionic (-soluble) concentrations are considered to represent the

bioavailable fraction of total mercury, with soluble fraction defined as the fraction able to

pass through a 0.45 ptm sieve. Partitioning coefficients are shown in Figure 4-2.











Complexed Complexed
Inorganic Mercury Organic Mercury
Methvlation
Washload % a / .i-ldiload
I L gK,,- 4,78 1 "I' I

ICS Soluble Soluble 1-
LogK, 1 19 I H| ,, .g2- y'"

Bedload Bedload
Logk, = o 3 I Demethylation LogKd = 0.24




Figure 4-2. State variables measured with MERC4. (Reprinted with permission from
Carroll, R.W.H., 1999. Simulation of Mercury Transport and Fate in the
Carson River, Nevada. Masters Thesis. University ofNevada-Reno, Reno,
Nevada. Figure 1.4, pp. 16)

Based on these considerations, the previous IFMTFM version was calibrated to

provide a best-fit to field observations of water column mercury and sediment

concentrations. This approach, however, is clearly not entirely representative of actual

field processes. The calibrated methylmercury function predicts methylmercury bank

concentrations over two orders of magnitude higher than the methylmercury bank data

collected under conditions of near-zero flow (Carroll, 1999). The samples, from October

1994, were collected high up on the bank at the estimated point of river contact during

spring floods. Examination of water column data shows that, unlike other systems,

increased discharge does not always cause a dilution in MeHg. Instead, an increase from

low to medium flow causes a significant increase in MeHg water column concentrations.

These results strongly suggest that given medium flows, bank erosion, and not diffusion,

acts as the principal pathway for MeHg into the water column. This trend of MeHg

increase with increasing flow begins to reverse when flows go from medium to high

conditions. The resulting dilution at higher flows suggests that the availability of MeHg









from erosion of the upper bank is limited, and this observation was used to justify the

inclusion of a threshold height got bank MeHg in the previous model version. This

limitation is not observed for inorganic mercury (Carroll, 1999).

Modeling MeHg concentrations as constant over time is not representative of field

conditions, which depend on microbe populations and therefore moisture levels. Mercury

methylation is stimulated if soils are fully saturated and anoxic conditions are achieved

(Gilmour and Henry, 1991; Porvari and Verta, 1994). Low flow conditions, which

characterize the Carson River during most of the year, do not result in inundation of

channel banks or the attainment of anoxic conditions. It is probable that high flow rates

are too infrequent and short in duration to allow anaerobic microbes to develop in the

upper portions of the bank sediments. Given these considerations, MeHg concentrations

within the channel banks are not expected to be vertically uniform; rather, MeHg

concentrations are expected to display a dependence on soil moisture conditions (Carroll,

1999). Flow history and the rates at which river stages are rising or falling will, in turn,

impart time-dependent and vertically variable MeHg concentrations within the channel

banks along the Carson River. This hypothesis is borne-out in soil samples collected at

Deer Run Road. Four soil samples were collected in triplicate from the bank profile at

this location, with sampling locations shown in Figure 4-3(A). Evaluation of sample

water content and percent methylmercury resulted in the relationship shown in Figure

4-3(B), which clearly shows a direct relationship between the percent of MeHg in the

bank profile and the soil moisture content (Carroll, 1999).

As mentioned previously, the previous version of IFMTFM utilizes temporally constant

MeHg bank concentrations and a threshold flow depth above which MeHg does not occur









in the channel banks, both of which were determined through calibration procedures. The

incorporation of the bank element into the IFMTFM attempts to remedy this and other

discrepancies between the model and physical observations; and reduce the number of

calibration parameters. Modeling efforts using the Finite Element Surface-Water

Modeling System: Two-Dimensional Flow in a Horizontal Plane (FESWMS-2DH) model

and analysis of water table data from the installed groundwater monitoring wells allowed

for the determination of a temporally-variable zone of saturation (as critical for

methylation) within the channel banks of the updated IFMTFM version. The zone of

saturation is a function of time history of channel stage, as computed within RIVMOD.

The saturated extent of channel banks is also a function of spatial location within the

model domain, resulting from variable channel geometry, and hence computed stage, for

any given flow magnitude. The creation of the new bank element type within IFMTFM

allows for moisture content within the banks will be continuously simulated and updated

as required throughout the model domain from CCG to Lahontan Reservoir.

Within the domains of RIVMOD/WASP5/MERC4, each bank element includes the

channel bank between adjacent stations (Carson River model includes 274 stations).

Recall that different channel geometry is assigned for each station within the input files.

This geometrical configuration inside each model segment (between adjacent stations) is

determined through linear interpolation of the properties of the two bordering stations.

Symmetry is specified for opposite channel banks. A single bank element is therefore

sufficient for each IFMTFM model segment.

Flow history must be tracked. The combined vertical extent of the capillary fringe

and the partially saturated zone will depend on whether river stage is rising or falling. A

































0.3

0.25

S0.2

S0.15

S0.1

0.05

0


-2.3 rm





-0.91 m


* ".


dry, closest to intermediate intermediate, 6" saturated
top deep
soil moisture content


Figure 4-3. Dependency of methylmercury concentration on soil moisture.
(A) Schematic of bank profile at DRR with sampling locations marked. (B) Percent
MeHg versus percent moisture content found in the bank profile at DRR. (Reprinted with
permission from Carroll, R.W.H., 1999. Simulation of Mercury Transport and Fate in the
Carson River, Nevada. Masters Thesis. University ofNevada-Reno, Reno, Nevada.
Figure 1.13, pp. 38)


smaller distance would be expected if the river stage is rising (bank not recently


inundated) and a larger distance if river stage is falling (bank recently inundated). This is


a by-product of the variable relationships between pressure head (Y) and moisture


content (0), based on whether a given soil is wetting or drying. This relationship is


commonly referred to as hysteresis and can be represented by scanning curves (Tindall


and Kunkel, 1999).


Sample Location
-----------------------------


fr









Two soil samples were collected along the channel bank at the location of the

groundwater wells during October 2002. While possible to obtain a steady-state estimate

for the size of the capillary fringe using these samples, incorporation of this value into the

IFMTFM would require the assumption of equilibrium attainment within bank sediments.

Given the nature of Carson River streamflow, equilibrium conditions will rarely be

reached (except perhaps during flood events). Ideally, data would be collected relating

the extent of the capillary fringe to the rates at which channel stage is rising or falling.

Realistically, this data collection is impractical and, given the spatial heterogeneity of

sediments within the Carson River system, unilaterally applying these functional

relationships to all bank sediments within the model domain is incorrect. Alternatively, it

will be kept in mind that a capillary fringe does exist, that MeHg concentrations within

channel bank sediments are likely underestimated, and, by neglecting this fringe, that the

final calibration to match observed MeHg water column concentrations will be

conservatively low. Because mass failure along channel banks is not a modeled erosion

process within IFMTFM, mercury enters the water column only through erosion or

diffusion from contacted (by stream stage) portions of the bank. Mercury concentrations

within the capillary fringe will only impact the model if these portions of the bank

become contacted in the immediate future. Relative to the contacted portion of the

channel bank, the extent of the capillary fringe is predicted to be small. With these

considerations in mind, attempts are only be made to track the saturated portion of the

channel bank elements.

Soil heterogeneity, the lack of available data regarding soil properties relating to

soil moisture, the presumed minimal impact of the capillary fringe on water column









mercury concentrations, and data analysis suggesting a nearly instantaneous response

time from channel to wells ultimately pushed modeling efforts away from the inclusion

of unsaturated zone modeling. The possibility of a variable time lag, from river stage

change to water table response, based on whether the bank sediments are wetting or

drying, however, was continually considered throughout the data analysis portion of this

project.

By tracking the size of the saturated zone over time as river flows change, the

updated IFMTFM begins to remove itself from the unrealistic assumption of static MeHg

concentrations within the channel banks and no MeHg above a given depth threshold, as

present in the previous model version. By allowing the model to run for a sufficient

length of time prior to generating used output, initial conditions (bank concentrations)

effectively wash out of the system, thereby removing the dependence of output on initial

conditions. The addition of the bank element allows for user specification of different

methylation and demethylation rates within the channel banks, as compared to the

channel bed. This is expected based on the higher percentage of fine particles within

channel bank sediments. Data has been presented to suggest that methylation is

dependent on moisture content and, specifically, the presence of anoxic regions within

the channel banks. Ongoing research at other institutions is seeking to determine these

rates as a function of soil moisture content and the time duration of saturation conditions

necessary for maximum MeHg production by microbes; preliminary results are used in

this thesis research.

As moisture content within the bank elements change over time, the concentrations

of methylmercury within the channel banks become dynamic, variable with respect to









time, spatial location within the model domain, and vertical location within the bank.

Variable concentrations are then used in the model to determine methylmercury amounts

introduced to the water column via erosion, diffusion, and advection (during receding

hydrographs) from the banks. Note also that while erosion of channel bank MeHg and

inorganic Hg(II) was previously simulated by the IFMTFM, diffusion of these chemicals

from channel banks is an addition to the code. The inclusion of bank diffusion could be a

significant contributor to MeHg concentrations in the water column under medium to low

flow conditions. Diffusion was previously only modeled from the bed sediments, which

include a much larger contact area for diffusion than do the channel banks. However,

mercury concentrations within the channel banks are observed to be two to three orders

of magnitude greater than the concentrations within the channel bed sediments.

Additional considerations were also needed to determine the impact of moisture content

history on methylation and demethylation rates. Although the rates are a function of

moisture content, the microbially-mediated nature of these chemical reactions does

suggest that response time for each microbial community should be considered and

somehow modeled. For example, how long does a portion of the bank need to be wet

before the observed net methylation rate reflects this moisture change? These

considerations are later discussed in chapter 7. Additionally noted is that water

temperature, which impacts the kinetics, is known by the model.

Final Considerations and Comparative Analysis

The completion of modeling efforts allows for an improved understanding of the

relationship between river stage change and water table. Based on the high degree of

spatial heterogeneity within the system, the simple relationships developed for the Scout

Camp reach are unlikely to precisely duplicate the field conditions observed elsewhere









within the Carson River system. The detailed FESWMS-2DH and groundwater data were

only tools used in developing the relationships between flow history and bank saturation

levels. The IFMTFM simulates the hydrodynamics within the system based on user-input

data from the USGS gages and through application of RIVMOD. RIVMOD is capable of

predicting the temporal and spatial variability of river stage. The water table observations

are not known to the IFMTFM model, as the groundwater wells were not installed during

periods of historical data collection (used for model calibration and verification).

Modified input data file structure and a new portion of FORTRAN code are added to the

IFMTFM to incorporate bank elements and dynamically compute and update MeHg

concentrations in the channel banks.

With the bank element concept presented to the reader, it is possible to move

forward to the model development and data analysis portions of this thesis. Following the

discussion of the data modeling and data analysis, the final component of this thesis

research is a comparative analysis to compare RIVMOD and FESWMS-2DH (SMS)

predictions for river stage and flow velocity as functions of flow magnitude. Ideally, the

results would be in perfect agreement. Regardless of the results, it must be realized that

the IFMTFM will still only have available to it the stage output from RIVMOD (a 1D

hydrodynamic model based on coarse, simplified channel geometry). However, the

potential use for higher resolution channel geometry data (for future model inclusion)

can be assessed.

Bank elements are used to model the spatial and temporal variability of bank

moisture conditions (over a range of flow conditions and scenarios) to determine bank

MeHg concentrations; these bank concentrations are then the determining factors for






62


water column concentrations and, ultimately, the availability of MeHg for uptake within

the aquatic food chain. The development and inclusion of bank dynamics within the

IFMTFM will assist in the modeling efforts for the mercury-contaminated Carson River

system, with the long-term goal of assisting in the determination of an appropriate

remediation strategy. Should additional data regarding methylation kinetics as they relate

to soil moisture content become available, the basic IFMTFM code structure needed for

its inclusion is in place following the completion of this thesis research.














CHAPTER 5
SURFACE WATER MODELING

Site Selection

The starting point for all field research is site selection. In the case of this study of

the Carson River, one of the goals is the quantification of the interaction between surface

water and groundwater. Months in advance of the first field expedition, it was decided

that this would be accomplished through a combination of surface water modeling and

analysis of data collected from near-channel groundwater wells. Site selection therefore

hinged on two critical elements: the ability to gain vehicular access to the site for the

survey equipment and the drill rig for well installation, and permission of the landowner

to install the wells. Additionally, the site needed to have the characteristic vertical banks

and a site layout conducive to and stable enough for getting a drill rig as close to the

channel as possible.

Much of the land along the lower Carson River is privately owned by cattle farmers

and agricultural interests, landowners with no real reason to justify access to their land

and not interested in supporting a water quality study. Fortunately, a portion of the lands

adjacent to the lower Carson River are public lands under the administration of Fort

Churchill State Historic Park of the Nevada Division of State Parks. On March 5, 2001,

accompanied by Mr. James Prida of the Fort Churchill State Park, a suitable site was

located at a Boy Scout Camp approximately 10 km downstream of the Fort Churchill

USGS Gaging Station (Figure 1-2). On March 8, 2001, three groundwater wells were

installed along the south bank of the Carson River at the Scout Camp field site. The three









wells were installed along a transect, perpendicular to streamflow direction, at distances

from the vertical channel bank of approximately 5, 25, and 100 feet. The 6-inch diameter,

PVC wells are cased to a depth of 20 feet. Pressure transducers monitor the position of

the water table within each well, data which is recorded by data loggers at 1-hour

sampling intervals.

Collection of GPS Survey Data

The channel geometry used in RIVMOD can be described as coarse, based only on

outdated USGS topographic maps. The construction of a surface water model required

more precise channel geometry data. The Leica GPS System 500 was used to conduct a

detailed bathymetric survey of the Carson River channel for a reach centered on the three

installed groundwater wells. The system utilized a base station and a rover unit. Survey

data was collected over two periods: March 6-9, 2001 and July 2-6, 2001.

The complete set of surveyed elevation data points is shown in Figure 5-1. Data

points could be divided into several classifications: transects (100 through 500), top of

vertical banks, water surface, channel thalweg, or general channel bed points. Surveyed

elevations are expected to be accurate to within +/- 3.0 centimeters (cm), with horizontal

accuracy closer to +/- 1.0 cm The total surveyed reach measured approximately 500

meters along the thalweg.

The bathymetric data collected was used as the framework for a surface water

model. As the raw surveyed data will later be refined, a few shortcomings of the survey

process are herein noted. Recall that GPS is based on travel times and distances from

receiver to at least four satellites. The large canopy of the cottonwood trees situated along

the channel banks often caused a loss of signal with the required satellites, resulting in

incomplete data along the top of the channel banks and effectively limiting the ability to











4503850
Cross-Section 500

4503800

S* Cross-Section 400 -- Flow Direction
g 4503750 **
z NIF

z 4503700 -

S4503650 Cross-Section 200
4503650 -- I


S4503600*
Cross-Section 300 Cross-Section 100

4503550 -*------


4503500
743800 743900 744000 744100 744200 744300 744400
Easting Coordinate (State Plane NAD 83, Nevada West)

Figure 5-1. Spatial location of survey points and cross-sections at Scout Camp site

collect precise elevation data for the floodplain. Vertical banks between two and three

meters in height additionally limited the ability to collect elevation data adjacent to the

base of the vertical banks. Observed streamflows during March ranged from 165 to 215

cfs, while those in July were between 8 and 9 cfs. Given that flows of interest are those

high enough to contact the channel banks (thereby causing erosion and introducing MeHg

to the water column), the water surface data collected in March is of particular interest for

surface water model calibration. However, the high flows at this time prevented the

acquisition of thalweg elevations along the channel, and thus prevented an analysis of any

relation between water surface slope and representative bed slope. For this reason, most

channel bed data was collected during the period of significantly lower flows. All data

collected are associated with measurement errors. This error is reported during the post-

processing stage. Unfortunately, some of the data points required multiple attempts at this









collection, as a result of errors outside of the acceptable range. These limitations, in

addition to time constraints, resulted in a surveyed reach of basically 500 meters,

including 5 complete transects. The spatial extent of the model domain is an important

component to SMS model development.

Surface Water Model Selection

Based on cost considerations, research, and consultation with the model developers,

it was decided that Surface Water Modeling System, Version 8.0 (SMS 8.0) would be the

modeling system of choice for developing a surface water model of the surveyed reach of

the Carson River. Within the SMS model, several modules are available. The module of

choice is FESWMS-2DH. The FESWMS model executable code, FLO2DH, is written in

the Fortran 77 computing language. FLO2DH was originally developed by the USGS

(Water Resources Division) and later modified by the Federal Highway Administration

of the U.S. Department of Transportation.

FESWMS-2DH applies the finite element method to solve the system of equations

that describes 2D depth-averaged flow in a horizontal plane. The system can be used to

simulate flow in rivers and floodplains and other surface-water bodies where vertical

accelerations are small in comparison to horizontal motion (Froehlich, 2000). The

hydrodynamic modeling code of FESWMS supports both super- and sub-critical flow

analyses, including area wetting and drying. Use of this model is justified because one is

not interested in the vertical variability of flow properties, but rather only the horizontal

distribution of flow quantities. The temporal variability of flow within the Carson River

channel also mandates the ability of the chosen model to account for wetting and drying

patterns within grid elements along the channel boundary. Within this document, the









surface water hydrodynamic model will be interchangeably referred to as SMS,

FESWMS, or FLO2DH.

Functionality of the FESWMS Model

Several steps were involved in the development of a functioning surface water

hydrodynamic model. GPS data (Figure 5-1) was used to generate a quasi-3D finite

element model of the reach of the Carson River immediately surrounding the Scout Camp

and the groundwater wells. Channel bathymetry data, from the field survey, was first

used to generate a 2D finite element network (or mesh) using the DIN2DH program. The

finite element network was then refined through node addition and deletion, elevation

modifications, and node spacing adjustments. With the model domain carefully planned,

the final step was to simulate water flow through the model domain. Though the

procedure is straightforward, the execution of this process for the Carson River model

proved far more complicated. Several months of research was placed into the

development of this model, and each step of model development will be reviewed in turn.

The primary equations governing the hydrodynamics are summarized by equations

5-1, 5-2, and 5-3 (Froehlich, 2000):

* Vertically-integrated momentum equation in the x direction:


O(HU) +s 1
+- arf HUU +(cos a, cos a )2 gH 2+-(,HUV)
at ax 2 ) y
(Eq. 5-1)
az 1 a(Hr ) a(HT~,) 0
+ cos a. gH b- oV + bx xY = 0
a P ax Oy


* Vertically-integrated momentum equation in the y direction:









O(HV) 0 1 a
(HD +- + fHVV+(cos a, os a')2 gH2 + (p HVU)
at 2 x
(Eq. 5-2)

+ cos a gH ab + 1DHU + r(H (Hr, = 0
S y p x


* Continuity equation


8H 8(HU) 8(HV)
H+ (H q (Eq. 5-3)
at ax y


where variables are defined as follows (units: [L] = length, [T] = time, [M] = mass):

H = water depth [ L ]
z b = bed elevation [ L ]
z s = z b + H = water-surface elevation [ L ]
U = depth-averaged velocity in the horizontal x direction [ L / T ]
V = depth-averaged velocity in the horizontal y direction [ L / T ]
Puv,Pvu,Pvv = momentum flux correction coefficients that account for the variation
of velocity in the vertical direction; coefficients result from the vertical
integration of the momentum transport equations and account
for the vertical variations of U and V
ax = arctan (az b / Cx)
Uy = arctan (8z b / y)
az = arccos (1 cos2x- cos2a)
g = gravitational acceleration [ L / T2
Q = Coriolis parameter [ T1 ]
p = water mass density, which is considered constant [ M / L3
bx, T by = bed shear stresses acting in the x and y directions,
respectively [ M / T2 L ]
Tsx, Tsy = surface shear stresses acting in the x and y directions,
respectively [ M / T2 L ]
TCxx, yx, yy = shear stresses caused by turbulence where, for example, cxy is the
shear stress acting in the x direction on a plane perpendicular to the y
direction [ M / T2 L ]
q = unit source (inflow) or unit sink (outflow) term [ L3/T / L2

These governing equations are solved by the FLO2DH program using the Galerkin finite

element method applied to the user-defined grid network. The method is carried out using









Newton-Raphson iteration and Gaussian quadrature integration for all elements

(triangular or quadrangular shapes defined by nodes at vertices and mid-side points)

within the model domain. FLO2DH is capable of simulating both steady and unsteady

(time-dependent) 2D (in a horizontal plane) surface-water flow to obtain depth-averaged

velocities and flow depths. The effects of bed friction and turbulent stresses are included,

as are, optionally, surface wind stresses and the Coriolis force (though the latter two are

not used here). The postprocessor program ANO2DH is used to present the results of

flow simulations either graphically or as reports. Graphic plots of velocity and unit-flow

vectors; ground-surface and water-surface level isolines; and velocity, unit flow, or stage

(water-surface elevation) time-history plots at a computation point can be created.

The principal gain from creation of this model is the generation of time-history

plots of river stage at the location of the groundwater well transect. An assessment can be

made as to the agreement between river stage over this reach as predicted by RIVMOD

and FESWMS-2DH; this will be discussed in the comparative analysis portion of this

document (chapter 8). Recall that the former (RIVMOD) is based on a generalized

channel shape and data interpolated from topographic maps, while the latter (FESWMS-

2DH) is based on precise data collected using GPS. The ultimate requirement for bank

elements is to predict moisture content variability within the channel banks, as a function

of temporally- and spatially-varying river stage, as predicted by the current RIVMOD

portion of IFMTFM. The FESWMS-2DH model was used to develop simple

relationships between flow history and bank moisture, but the FESWMS model is not

directly incorporated into the IFMTFM; the simple relationships are rather applied to

river stage as predicted by RIVMOD.









Creation of Scatter Point Data Set

Data points for a finite element mesh can be generated directly from topographic

data, such as a list of survey points within a text file, imported into SMS. The Leica GPS

software allows for the extraction of data point latitudes, longitudes, and elevations in the

form of a simple text file. In this instance, the text file was a compilation of the survey

data collected during March and July of 2001. Survey point data from the Leica output

text files were first imported into Excel. Using examples of .xyz files included in the

FESWMS documentation, the raw survey data was copied and manipulated within Excel

to match the required input file format for FESWMS.

Pre-Processing of Data Prior to Mesh Construction

The FESWMS mesh was constructed by making use of the available survey data.

Before mesh creation, a moderate degree of pre-processing was required to account for

shortcomings of the data set. Limited data downstream of cross-section 100 suggests that

this cross section will serve as the downstream boundary for the surface water model.

Surveyed points downstream of this cross-section were therefore deleted. Erosion and

deposition processes result in dynamic bed forms (high and low spots) and seasonal

changes to exact thalweg location. In the interest of consistency (and with the most

complete data available in July), survey points along the channel bed collected during

March (water surface elevations) were removed from the scatter data set. To preserve the

valuable detail provided from the cross-section surveys, however, the data points from

the four downstream cross-sections collected in March are retained. Duplicate data points

collected at nearly the same spatial location, as evidenced through visual inspection, were

selected and deleted. While the software does include options for duplicate point

removal, these means could not be used because of the spatial proximity of desirable









points at the top and base of steep banks. One is generally seeking a uniform nodal

spacing in the range of 3 to 10 meters. Finer resolution is needed in areas where the

element slopes are steep, such as along the banks, and lower resolution is adequate for

flatter regions, or where survey data is less extensive. The identification of points to be

removed was tedious as each point must be identified by coordinates only, but trial and

error indicated this to provide the best starting point for mesh generation, as these points

determine the mesh boundary and the first cut for the mesh layout.

With the point set modestly trimmed down, the Excel file was saved as a text file,

and the file extension was changed to .xyz. The survey points are now able to be read into

FESWMS using the 'File / Open' command. Within the File Import Wizard dialog box,

the option was selected to open file as mesh and the coordinate systems are set to match

the survey data (Horizontal System: State Plane NAD 83, State Plane Zone: Nevada

West 2703, Vertical System: NAVD 88, Units: meters) before the file was opened. The

data points from the file were converted to mesh nodes, points representing locations of

known position and elevation. The key components to a finite element mesh are the

elements themselves, which connect the nodes to define the flow area, or model domain.

The network of scatter data points (the dots) and the associated mesh boundary is

shown in Figure 5-2. For reference, the elevation contours created from this initial data

set have also been overlain. All SMS-generated figures are in color, with the scale and

type of plotted contours (water surface elevation, water depth, ground elevation /

bathymetry, or velocity) indicated by the legend. A few items are of particular note.

Vertical banks are poorly represented due to the inability to properly survey the base of

the banks, or the top of the banks, in certain instances. Evidence of the channel thalweg






















elevation : *
1254.80
- 1254.30 ', L ... I

1252.80 .
1252.30 I
v ^ '"- : \%_/
1251.80 ** -
1251.30
1250.80 -
- 1250.30 "' "-" --

Figure 5-2. Original survey points after import into SMS scatter module

(or low-flow inner channel) is present, but many of these points are disjointed.

Additionally, the linear interpolation scheme of the model does not necessarily connect

those points which should logically be connected. The plethora of points along the five

channel cross-sections is also very clear, as is the paucity of points along some portions

of the channel banks and bed. Clearly, nodal spacing is also far removed from uniform.

The resolution of these shortcomings is discussed at length in Appendix A.

Mesh Construction and Editing as Discussed in Appendix A

The mesh construction and editing process is discussed in Appendix A (separate

from the main body of text) not because it was not a critical part of this thesis, but rather

because the material is not needed for the reader to understand and appreciate the results

of this research. The material is available for the reader's reference for three main

reasons: to provide proof that mesh construction was completed under a carefully planned

set of guidelines, to provide explanation for the mesh's limitation regarding the modeling









of overbank flows and the simulation of dynamic streamflows, and to justify the time

spent getting the model to simulate Carson River hydrodynamics.

The development of a functioning model based on this set of scattered survey data

points proved to be far more labor-intensive task than initially believed. Model

development required several months before the model was able to simulate streamflow

conditions ranging from bank full levels to levels at which flow was confined to the

channel thalweg. Efforts were made to retain channel morphology information contained

within the survey data set. Attempts to model stage levels above bank full and on the

floodplain proved unsuccessful. Several problem areas were encountered as next listed:

model parameterization, including element wetting and drying and specification of

boundary conditions; steep channel banks, including modified mesh boundaries and slope

transitions; irregular channel morphology, as result from erosion and deposition

processes; drying of elements internal to the wet / dry boundary; jagged wet / dry

boundaries; element re-wetting; supercritical flow conditions along boundary

nodestrings; nodal velocity and water surface divergence; non-uniform node spacing and

limited data availability for some portions of the mesh, resulting in poor element

construction; and approaches for attaining model convergence. The mesh was constructed

to accurately portray the geometry of the channel, as based on a combination of survey

points, field measurements of channel bank dimensions, photographs, and personal

recollection. The desired end-product was a model able to wet and dry as would be

observed in the field. When the mesh editing process was complete, the FLO2DH

hydrodynamic model was able to simulate element drying from bank-full to flows

confined to the channel bed.









Summary of Mesh Editing

No amount of written explanation can explain the frequent frustration associated

with getting this model to simulate the range of flows from bank-full to nearly zero. Any

change in mesh geometry, regardless of how trivial, required the model to be initialized

with all elements submerged. Water surface lowering needed to be performed gradually

to allow the time for model convergence. Iteration after iteration, attempts were put forth

to identify as many potential problem areas as possible, but it was never known if the

problem was corrected until the water surface was able to be lowered beyond a previous

threshold. It remained uncertain if this was the last round of mesh editing or if many

more divergence problems awaited discovery. Eventually, however, the mesh was

completed. The water surface was able to be lowered to minimum levels at which flow

could no longer pass through the model domain, due to interior elements along the

thalweg becoming dry. After all the frustration, repetition, and uncertainty, the sense of

satisfaction and accomplishment that is worthy of note. The final edited mesh

(bathymetry) is shown in Figure 5-3. The well transect is indicated in the figure. For

reference, the minimum flow levels able to be modeled are also shown in Figure 5-4. The

flow magnitude at this condition is not discussed here as this early result was based on a

0.2-meter head gradient between the upstream and downstream boundary condition

nodestrings, a gradient originally chosen for no other reason than to be conservative.

With the model able to simulate element drying under these conditions, the model was

predicted to be stable for any larger head gradient and the associated higher mean

velocity. Admittedly, most of the required mesh editing was due to the manner in which

the mesh was constructed, that being from a random network of survey points. While the

model might have performed better had less detail been included, this would also have










significantly altered the reliability of the model results. After all, the modeled reach of the

Carson River is only 500 meters in length, and the goal was to accurately simulate the

hydrodynamics (and thus river stage) down to low flows within the channel thalweg,

which necessarily requires a high degree of model detail.












River Cross-Section
at Well 1





1254.20
1253.63
1253.05
1252.47
1251.90
1251.32
1250.75
1250.17


Figure 5-3. Bathymetry of Scout Camp reach after mesh editing found in Appendix A








76


















epth:130.000
2.50
2.15
1.79
1.43
1.08
0.72
0.37
0.01





Figure 5-4. Minimum flow levels able to be modeled, subject to a 0.2-meter head
gradient














CHAPTER 6
WATER TABLE RESPONSE TO STREAMFLOW VARIABILITY

Linking Surface Water Modeling and Groundwater Data

When the SMS model was selected as the surface water model for this project, two

distinct end-products were envisioned. The first was to develop a simple relationship to

relate stage in the channel to water table level within the bank. This relationship was then

to be incorporated into a new IFMTFM bank element functionality to predict saturation

conditions within the channel bank immediately adjacent to the channel, as a function of

flow depth and channel geometry known by RIVMOD. The second end-product, to be

discussed in chapter 8, was a comparative analysis between RIVMOD predictions and

SMS predictions. For the modeled Scout Camp reach, a comparison between the two

models was sought for flow depths and mean velocity to see if the simplified channel

geometry and ID nature of RIVMOD results in consistently high or low predictions

relative to a more detailed and calibrated model.

Data Availability

A GPS survey was conducted to provide channel bathymetry input data to the SMS

model, which was used to model Carson River streamflows and predict river stage. After

site selection in March 2001, the survey operation began. With the groundwater wells

sited, the intention was to survey channel cross-sections both upstream and downstream

of the well transect, with roughly equivalent survey areas in both directions.

Complications arose with loss of satellite lock caused by cottonwood trees, vertical

banks, and, at times, less-than-ideal satellite configurations. Further complications arose









when post-processing indicated points with unacceptable error values, which would have

to be surveyed again. After two channel survey operations in March 2001 and July 2001,

five channel cross-sections and approximately 500 meters of channel length were

surveyed. Three groundwater wells are situated along a transect, approximately 275

meters downstream of the upstream boundary condition of the mesh, as indicated in

Figure 5-3.

The SMS mesh was constructed (Appendix A) to predict the relationship between

streamflow and river stage, specifically the stage adjacent to the vertical bank at the well

transect. Three gages along the Carson River monitor streamflow levels. The RIVMOD

module in IFMTFM incorporates dynamic streamflow values from the Carson City and

Fort Churchill gages to perform hydraulic routing. Upstream irrigation withdrawals are

accounted for using the Fort Churchill streamflow values. With no known withdrawals

downstream of Fort Churchill, RIVMOD is capable of routing flows and predicting

average streamflows for all RIVMOD segments downstream of Fort Churchill.

Generation of Boundary Condition for FESWMS Surface Water Model

To develop a relationship between river stage and water table level, the river stage

needed to be accurately modeled within the FESWMS module of SMS. Following mesh

creation, editing, and refinement; but before simulating the hydrodynamics through the

mesh reach, boundary conditions must be specified. The FLO2DH model requires an

upstream and a downstream boundary condition.

During the mesh construction process (Appendix A), it became evident that the

finite-element mesh was not suitable for the simulation of dynamically rising and falling

flows. The decision was made to use head boundary conditions at both the upstream and

downstream extents of the model spatial domain. Nodestrings along the boundaries and









strategically located at cross-sections within the mesh interior were used along with the

model solutions for flow depth to identify the flow through any given cross-section.

Water surface elevations from a FESWMS model solution are, therefore, relatable to

streamflow through the model domain.

Within FESWMS, the head boundary conditions are incrementally lowered to

generate a series of water surface solutions. The FESWMS module can stably operate

with a head or flow upstream boundary condition, but the model requires a head

downstream boundary condition. Because the Scout Camp site is not located at a gaged

location, a rating curve at the downstream boundary was not available. The head gradient

between upstream and downstream boundary conditions must, instead, be specified and

held at a constant value, based on the lack of stage data for the downstream boundary and

a lack of data for the variability of water surface slope with streamflow magnitude, as

later discussed. Though an upstream head boundary is used in the surface-water model,

the flow hydrograph at the upstream boundary channel cross-section must also be known

in order to later link surface water modeling of river stage to the corresponding water

table elevations from the wells within the channel bank.

Streamflow routing: testing the validity of instantaneous translation

Water table data collected via the groundwater wells is a time series of water table

elevation (NAVD 88). Flow in the channel also varies over time, as evidenced by the

hydrographs of USGS gaging stations at Carson City and Fort Churchill. This flow must

be specified to develop a correlation between water table and river stage. Two options to

obtain this time series were considered. First is the possibility of an evapotranspiration

(ET) forcing function. Under this scenario, streamflow would simultaneously rise and fall

at all locations along the Carson River. Peak flow would occur everywhere at the same









time of day. Differences in streamflow magnitude with distance are the result of irrigation

withdrawals and tributary additions. Given the arid environment and the prevalence of

transpiring cottonwood trees along the entire river length, this seemed a viable

possibility. Streamflow withdrawals for irrigation are believed to be nonexistent

downstream of the Fort Churchill gaging station. If the ET forcing function was indeed

present, so that streamflow oscillation is solely due to flow withdrawals, then the

streamflow time series at Fort Churchill could be instantaneously translated (no

consideration of travel time) and used as the upstream boundary condition for the Scout

Camp reach. This instantaneous downstream translation by a distance of 10 km may not

seem intuitively correct, as a reasonable velocity of 0.5 m/s would translate to a 6-hour

travel time. However, it would simplify the effort to obtain the streamflow time series at

the upstream boundary, and the option was investigated.

To validate or invalidate the assumption of instantaneous translation and an ET

forcing function, streamflow data from three Carson River gages were compared for the

period of available well data (March 2001 through September 2002). The three USGS

gages, Carson City Gage (CCG), Deer Run Road Gage (DRR), and Fort Churchill Gage

(FCH), record streamflow at 15-minute intervals. This set of data was obtained from the

USGS with the assistance of USGS Hydrologist Jim Crompton in Carson City, Nevada.

The streamflow time series are graphed in Figure 6-1 (2001) and Figure 6-2 (2002). The

data is separated by calendar year. A logarithmic scale is used to better capture the high

degree of streamflow variability over this period of interest. Streamflow data available at

15-minute resolution should capture the diel (within day) variability at these respective

locations.


















10000





1000





2 100
E


10
0

C.


0 1 I , l
3/7/2001 4/6/2001 5/6/2001 6/5/2001 7/5/2001 8/4/2001 9/3/2001 10/3/2001 11/2/2001 12/2/2001 1/1/2002
date and time


Figure 6-1. 2001 USGS streamflow gage comparison: 15-minute data


l 0 1 1 1 1 1
1/1/2002 1/31/2002 3/2/2002 4/1/2002 5/1/2002 5/31/2002 6/30/2002 7/30/2002 8/29/2002 9/28/2002
date and time


Figure 6-2. 2002 USGS streamflow gage comparison: 15-minute data













1600 ---
-Carson City Gage
1400 -- i -- -- Deer Run Road Gage





400
200 ----------------------------------------- ---o--urc-i----ChuhllGa






200











date and time


Figure 6-3. May 2002 Carson River streamflow at USGS gages

The complete set of data provides evidence of a travel time for flow peaks from

upstream gaging stations to Fort Churchill. A more noticeable time lag is evident for

DRR to FCH as compared to CCG to DRR, as reasonably expected given the 6-km

versus 52-km distance between the respective stations. A subset of the data, for May

2002, is used to illustrate this point, appearing in Figure 6-3. Note the diel variations in

streamflow. The daily peaks do not occur consistently at the same time of day, but rather

occur at times dictated by travel time from upstream. During times of year when

streamflow is primarily due to snowmelt, higher melt rates would be observed during the

early afternoon when the day is at its hottest. As the melt rates slow overnight, the result

is a diel streamflow variability which will then characterize the observed streamflow

further downstream. Despite the arid climate, the diel streamflow variability caused by

ET is clearly overshadowed by the influence of travel time. The observed streamflow









reduction from CCG to FCH is due to irrigation withdrawal; no known irrigation ditches

or tributaries occur from Fort Churchill to the study reach at the Scout Camp. The Scout

Camp site is approximately 10 km downstream from FCH, so a lag time is expected

between these two locations. Fort Churchill streamflows will need to be hydraulically

routed to create the upstream boundary condition for the SMS model. The assumption of

instantaneous translation would have the effect of artificially impacting the observed time

lag between modeled river stage (a function of streamflow) and water table oscillations.

In the extreme case of rapidly rising flow and rapid response time between channel and

bank, the assumption of instantaneous translation could presumably cause the well

response to precede the river stage oscillation. This approach was discarded.

Streamflow routing: hydraulic routing using RIVMOD

Streamflow was instead routed using RIVMOD. RIVMOD performs hydrodynamic

computations using a 60-second time step. Being only interested in the hydrodynamic

output from IFMTFM, there was no need to accurately simulate reservoir stage, which

impacts the mercury computations. This alleviated the need to obtain and process

15-minute data for the USGS gage in Lahontan Reservoir at the dam (LRD). A time

series for flow must still be input to RIVMOD at LRD, but the values were simply set

equal to the time series for flow at FCH. This approach ensures that backwater conditions

from the reservoir do not impact the hydraulic routing from Fort Churchill to the

reservoir, between which the Scout Camp is located.

Within IFMTFM, streamflow values input to RIVMOD are average daily flows. To

be consistent with the 1-hour interval at which water table data is collected by the well

data loggers, a 1-hour time interval was selected for streamflow input to RIVMOD for

the hydrodynamic routing. To create the RIVMOD input data files, macros were written









to extract hourly flow data from the smaller 15-minute data sets. Data from CCG and

FCH were extracted, and, as mentioned, flows at LRD are taken as equal to those at FCH.

Inflow to the reservoir from the Truckee Canal was modeled as zero. Care was taken to

ensure that the flow values at the respective stations within any given input file row

corresponded to the same point in time, as needed to compensate for the intermittent data

gaps within the 15-minute data files. Flows lower than 8.0 cfs were changed to 8.0 cfs, as

this value represents the minimum allowable flow input for RIVMOD within IFMTFM.

Periods of lower flow are therefore not accurately simulated by RIVMOD. Before

simulating real-time hourly streamflow data, RIVMOD was run at steady state for a

period of 10 days to create a valid initial solution throughout the RIVMOD model

domain from CCG to Lahontan Reservoir. Hydrodynamics were then simulated for the

period from March 2001 through August 2002 using hourly gage data, and the solution

time series for flow are output, to a text file, for RIVMOD segments 148 through 155

(segments near Scout Camp reach). It is also noted that two separate hydrodynamic

simulations were necessary because of file size limitations in RIVMOD; a 30 day overlap

was used to ensure flow continuity when the simulation results were later combined.

Latitude and longitude of all survey points are known from the survey data set in

Nevada West State Plane Coordinates. RIVMOD segments have previously been

recorded on a USGS topographic map, which utilizes the 1927 North American Datum

(NAD 27) as the horizontal reference. Within SMS, the mesh coordinates (surveyed date

points) are easily converted to an alternate coordinate system from the Edit menu. This

feature was used to identify the NAD 27 coordinates corresponding to the upstream and

downstream boundary conditions, as well as the location of the well transect. The









upstream and downstream boundary conditions are located within RIVMOD segments

151 and 153, respectively. The groundwater well transect is approximately centrally

located within RIVMOD segment 152. Segment spacing for model segments 144 through

240 is equal to 250 meters.

RIVMOD flow computations were used to construct the flow time series at the

upstream boundary condition of the FESWMS model. With no significant inflows or

flow withdrawals downstream of Fort Churchill, hydraulic routing results in hydrograph

attenuation and delay due to channel resistance. RIVMOD does not model hydrograph

attenuation due to storage capacity within channel banks. RIVMOD is able to smooth the

FCH hydrograph, which can oscillate significantly during periods of rapidly changing

streamflow or under low flow conditions (because of inherent gage limitations). The

hydrograph input into RIVMOD uses hourly time steps. The routed hydrograph output

from RIVMOD is extracted at hourly intervals, but all RIVMOD computations are

completed at 60-second time steps.

Groundwater Well Data

General Well Information

Three groundwater monitoring wells were installed in March 2001, approximately

centrally located within the surveyed Scout Camp reach. The 6-inch diameter, 20-foot

deep wells are located along the south streambank at distances from the vertical bank of

approximately 5 (Well 1), 25 (Well 2), and 100 (Well 3) feet. The three wells are situated

along a transect perpendicular to streamflow direction. Pressure transducers situated at

the base of each well monitor the water table level, with data continuously recorded by

data loggers. With the notable exception of the middle well, continuous water table data

is generally available dating back to the time of well installation.









Data Limitations

Although well installation was completed on March 9, 2001, time was allowed for

the water to clarify prior to data logger insertion. Collection of water table data began on

March 12, 2001, after the March 2001 field survey work was completed. Water table

elevations, relative to the NAVD 88 vertical datum, are measured in each well at hourly

intervals. The probe installed in Well 2 failed to function properly, and this data set is

therefore unavailable. After the failure of this probe, the remaining two probes were also

removed on May 29, 2001 for preventative maintenance, to ensure no similar problems in

the future. The probes were unable to be returned to the wells before the field work

period in July 2001. For this reason, a data gap appears in the water table data sets for

Well 1 and Well 3 between May 29 and July 5, 2001. During the reinstallation process,

the cable for holding up the probe in Well 1 was insufficiently stabilized. Duct tape failed

to hold the looped cable in place, and the probe weigh resulted in the probe falling to the

base of the well. As this process occurred gradually, a uniform correction is unable to be

applied to the well data. Probe slippage is evident in the well data for this time period, as

shown in Figure 6-4, and was verified by well inspection in October 2002 (note the

sudden rise in water table, inconsistent with water table oscillations observed in Well 3,

associated with the increased water pressure felt by the transducer as it moved down the

well). With the probe slippage complete and the probe at the base of Well 1, the slip

distance was calculated and applied as a uniform correction to the water table data for

Well 1 after August 1, 2001. The 2001-2002 water table elevations in Figure 6-4 are

plotted relative to the thalweg elevation at the well transect (1252.01 meters (NAVD 88)

set as point of zero-reference); additionally included for reader reference are the top of

bank elevation (1254.15 meters), the top of casing elevation for Well 1 (1253.95 meters),