Selenium removal processes from Great Salt Lake, Utah: estimating sedimentation and verifying volatilization fluxes

Removal processes for selenium (Se) from the Great Salt Lake, Utah (GSL) are investigated by estimation of sedimentation flux from analysis of lake cores and direct measurement of volatilization flux for comparison to a predictive model. To estimate Se removal by sedimentation, lake sediment cores were used to delineate qualitative sedimentation regions, estimate mass accumulation rates (MAR), and determine sediment 210 226 7 137 Se concentrations. Quantifiable MAR results from analysis of Pb, Ra, Be, and Cs activity in 8 deep cores ranged from 0.010 to 0.049 g/cm2/yr. Contemporary sediment Se concentrations from the upper 2 cm of each deep core ranged from 0.79 to 3.12 |ig/g. A representative MAR and Se concentration was assigned based on deep cores located within each sedimentation region. Coupling these with region area, mean annual Se removal by sedimentation was estimated to be 520 Kg/yr within a range of uncertainty between 45 and 990 Kg/yr. Volatilized Se from the water surface was contained and concentrated using a floating emission isolation flux chamber (St. Croix Sensory, Inc.) and captured in a cryogenic finger trap. These samples were taken concurrently with volatile Se concentration, wind velocity, and surface water temperature for input into the Se flux predictive model. Direct flux measurements under controlled laboratory conditions in which Se concentration varied and was independently verified suggest that 10% of actual flux was captured by the direct measurement. After correction for background and 10% measurement inefficiency, measured fluxes approximated, but were generally higher than predicted fluxes. The specific cause of this discrepancy is unclear, but the correspondence is strong enough with the limited number of direct measurements that a correction of the predicted flux is not warranted. The results of these two investigations compared to estimated loadings of Se suggest that volatilization, not permanent sedimentation, is primary removal process for Se from the GSL.

SELENIUM REMOVAL PROCESSES FROM GREAT SALT LAKE, UTAH: ESTIMATING SEDIMENTATION AND VERIFYING VOLATILIZATION FLUXES by Wade Austin Oliver A thesis submitted to the faculty of The University of Utah partial fulfillment of the requirements for the degree of Master of Science in Geology Department of Geology and Geophysics University of Utah May 2008 in III Copyright © Wade Austin Oliver 2008 All Rights Reserved THE U N I V E R S I T Y OF UTAH G R A D U A T E SCHOOL of a thesis submitted by Wade A. Oliver This thesis has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. Chair:/William P. Johnson D. Kip Solomon 4-9-c>s /U L. David L. Naftz UNIVERSITY GRADUATE SCHOOL SUPERVISORY COMMITTEE APPROVAL \ illiam y If- 0- o~ THE U N I V E R S I T Y OF UTAH G R A D U A T E SCHOOL FINAL READING APPROVAL To the Graduate Council of the University of Utah: I have read the thesis of Wade A. Oliver m [ts f m a \ form and have found that (1) its format, citations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the, supervisory committee and is ready for submission to The Graduate School. g - ? J - Q * . Date Willia^i P. Johnson Approved for the Major Department (Sue UJ. Pel Erich U. Petersen Chair/Dean Approved for the Graduate Council r. rpC^p David S. Chapman! Dean of The Graduate School UNIVERSITY GRADUATE SCHOOL APPROVAL I have read the thesis of -----------W--a-d-e -A-. -O-l-iv-e-r ----------- in its final form th Date Chair: Supervisory Committee Approved for the Major Department Chapman ABSTRACT Removal processes for selenium (Se) from the Great Salt Lake, Utah (GSL) are investigated by estimation of sedimentation flux from analysis of lake cores and direct measurement of volatilization flux for comparison to a predictive model. To estimate Se removal by sedimentation, lake sediment cores were used to delineate qualitative sedimentation regions, estimate mass accumulation rates (MAR), and determine sediment 210 226 7 137 Se concentrations. Quantifiable MAR results from analysis of Pb, Ra, Be, and Cs activity in 8 deep cores ranged from 0.010 to 0.049 g/cm2/yr. Contemporary sediment Se concentrations from the upper 2 cm of each deep core ranged from 0.79 to 3.12 |ig/g. A representative MAR and Se concentration was assigned based on deep cores located within each sedimentation region. Coupling these with region area, mean annual Se removal by sedimentation was estimated to be 520 Kg/yr within a range of uncertainty between 45 and 990 Kg/yr. Volatilized Se from the water surface was contained and concentrated using a floating emission isolation flux chamber (St. Croix Sensory, Inc.) and captured in a cryogenic finger trap. These samples were taken concurrently with volatile Se concentration, wind velocity, and surface water temperature for input into the Se flux predictive model. Direct flux measurements under controlled laboratory conditions in which Se concentration varied and was independently verified suggest that 10% of actual flux was captured by the direct measurement. After correction for background and 10% measurement inefficiency, measured fluxes approximated, but Se concentrations. Quantifiable MAR results from analysis of 21 0Pb, 226Ra, 7Be, and 137CS activity in 8 deep cores ranged from 0.010 to 0.049 g/cm2/yr. Contemporary sediment Se concentrations from the upper 2 cm of each deep core ranged from 0.79 to 3.12 ~g/g. A representative MAR and Se concentration was assigned based on deep cores located within each sedimentation region. Coupling these with region area, mean annual Se removal by sedimentation was estimated to be 520 Kg/yr within a range of uncertainty between 45 and 990 Kg/yr. Volatilized Se from the water surface was contained and concentrated using a floating emission isolation flux chamber (St. Croix Sensory, Inc.) and captured in a cryogenic finger trap. These samples were taken concurrently with volatile Se concentration, wind velocity, and surface water temperature for input into the Se flux predictive model. Direct flux measurements under controlled laboratory conditions in which Se concentration varied and was independently verified suggest that 10% of actual flux was captured by the direct measurement. After correction for background and 10% measurement inefficiency, measured fluxes approximated, but were generally higher than predicted fluxes. The specific cause of this discrepancy is unclear, but the correspondence is strong enough with the limited number of direct measurements that a correction of the predicted flux is not warranted. The results of these two investigations compared to estimated loadings of Se suggest that volatilization, not permanent sedimentation, is primary removal process for Se from the GSL. v fluxes . v This thesis is dedicated to P.G. aanndd BB..DD.. iinn ssiinncceerree aapppprreecciiaattiioonn ooff tthheeiirr lloovvee aanndd ssuuppppoorrtt.. TABLE OF CONTENTS ABSTRACT iv LIST OF TABLES viii LIST OF FIGURES ix Chapter 1. INTRODUCTION 1 2. ESTIMATING SELENIUM REMOVAL BY SEDIMENTATION 3 Abstract 3 Introduction 4 Methods 5 Results 16 Discussion 45 Conclusion 50 3. VERIFYING SELENIUM REMOVAL BY VOLATILIZATION 51 Abstract 51 Introduction 52 Methods 53 Results and Discussion 61 Conclusion 71 Appendices A. CORE TRACE ELEMENT CONCENTRATIONS 72 B. PURGE AND CRYO-TRAP SYSTEM 74 C. FLUX CALCULATIONS 78 REFERENCES 83 ........... ......... ...... .......... ...................................................... .. .... .. ..... . ..... ............ ..... . . ...... .................. ... .. ...... .......... .................................................................................. ..... .. ...... ..... ............. .. .......................... ..... ............... .................... ... . . ....... ....... .. ............ .... ......... .... ........ ......................... ....................................... .................... ..... ...... ........ .. .......... ....... ......... ....... .......... ......... ....................... ...... . . .... ........ ... .............. ... ............... ....... ....................... .... . .. ................. ... .............. .... ......... ........ .................... ......... ........ .......... ...... ............ . . .. .......................... ... ...... ... .. ...... ... ....... ..... ...... .. .......... ............. ..................... ..... ...... . .... .. ................................... ............... ... ....... ....... .......... .................. ...... ................... ........ ............ ....... . .. ........ .................................. ... .... ........... ............ ......... .... ............ . ................ . ......... . .. .... ...... .... .......... .. ............................... . .... SySTEM ....... ..... .. .. ... ........... .... ... . .. ...... ..... ......................... ... ...... .......... ...... ................. ...... ....... .... .. ...... ........... . . .......................... ...... .... ..... ........ LIST OF TABLES Table Page 1. Quality control summary 11 2. Core mass accumulation rate results 17 3. Salinity corrected deep core Se concentrations 18 4. Sedimentation region and mean Se concentration in cores 27 5. Mean 0-2 cm Se concentration, MAR, area, and Se mass removed annually within each sedimentation zone 29 6. Total zone relative standard deviation determination 36 7. Mean and range of estimated annual Se removal by sedimentation 36 8. Results of site 2267 shallow sediment trap analyses 41 9. Results of site 2565 deep sediment trap analyses 41 10. Results of site 2565 shallow sediment trap analyses 42 11. Results of site 3510 deep sediment trap analyses 42 12. Results of site 3510 shallow sediment trap analyses 43 13. Results of measured volatile Se fluxes under controlled laboratory conditions with variable sweep rate and sweep gas composition 60 corresponding 15. Results of attempted flux recovery test 66 .................................................................. .................................................... ........................................ ........................... annually within each sedimentation zone ............................................... 29 6. Total zone relative standard deviation determination ................................... 36 7. Mean and range of estimated annual Se removal by sedimentation ................. 36 8. Results of site 2267 shallow sediment trap analyses .................. ................. 41 9. Results of site 2565 deep sediment trap analyses ....................................... .41 10. Results of site 2565 shallow sediment trap analyses ................................... 42 11. Results of site 3510 deep sediment trap analyses ...................................... 42 12. Results of site 3510 shallow sediment trap analyses ................................... 43 13. Results of measured volatile Se fluxes under controlled laboratory conditions with variable sweep rate and sweep gas composition .................... 60 14. Measured and predicted volatile Se fluxes with corresponding environmental parameters ................................................................. 62 .................................................. LIST OF FIGURES FIGURE Page 1. Great Salt Lake sampling locations 6 2. Schematic of sediment traps for shallow site 2267 14 3. Schematic of sediment traps for deep sites 2565 and 3510 15 4. Selenium concentration profile at site 3510 19 5. Cobalt concentration profiles at sites 3510, 2565, and 2267 21 6. Copper concentration profiles at sites 3510, 2565, and 2267 22 7. Shallow core results overlain on Holocene isopach contours 23 8. Qualitative sedimentation zones based on Holocene isopach contours 9. Standard deviations as a function of the number of randomized MAR values assessed 33 10. Sedimentation and Se sedimentation fluxes at shallow site 2267 38 11. Sedimentation and Se sedimentation fluxes at deep site 2565 39 12. Sedimentation and Se sedimentation fluxes at deep site 3510 40 13. Se concentration chronology based on MAR in cores at sites 3510, DD-C, and DD-1 48 14. Se concentration chronology based on MAR in cores at sites DD-L, DD-Q, and DD-R 49 15. St Croix Sensory, Inc. Emission Isolation Flux Chamber during sample collection on calm day 56 ........................................................ .................................... ........................... ............................................ .19 .......................... ........................ ........................ and shallow core linear sedimentation rates ............................................ 25 MAR values assessed ...................................................................... 33 10. Sedimentation and Se sedimentation fluxes at shallow site 2267 .................... 38 11. Sedimentation and Se sedimentation fluxes at deep site 2565 ........................ 39 12. Sedimentation and Se sedimentation fluxes at deep site 3510 ....................... 40 13. Se concentration chronology based on MAR in cores at sites 3510, DD-C, and DD-I. ................................................................... 48 .................................................................. ............................................................ 16. Diagram of temperature-controlled cryo-focusing system for collection of volatilized selenium from GSL 58 17. Relationship between measured and Estuarine model-predicted volatilization rates from the GSL 65 18. Relationship between measured and Estuarine model-predicted volatilization rates under controlled laboratory conditions 68 19. Relationship between measured and Estuarine model-predicted volatilization rates from the GSL after inefficiency correction 69 20. Schematic representation of the volatile selenium cryo-focusing trap collection system 75 21. Calibration curve for dimethyl selenide using the purge and trap system 77 GSL. .......................................... GSL. ................... . ...... ....................... ...... ....... ...... .......... ..... ...................... ................. ... ........ ......... ............ . .......... ......... .......... x CHAPTER 1 INTRODUCTION The Great Salt Lake (GSL) has been viewed as a "self-cleaning system." As a terminal lake, evaporation is the only output for inflowing water in the GSL. Trace metals that enter the lake undergo geochemical precipitation and removal from the system by sedimentation (Tayler et al., 1980). Geochemical precipitation as metal sulfides is promoted in the anoxic deep brine layer. The deep brine layer underlies the deepest portions of the lake. Precipitation of metal sulfides is driven by sulfate reducing bacteria that populate the deep brine layer (Tayler et al., 1980). To date, sedimentation has been viewed as the primary removal mechanism for trace metals and other contaminants in the GSL. The predominant view that sedimentation in the lake is largely permanent is inconsistent with the lake's vast surface area relative to volume (shallow depth). Wind-driven mixing in the water column may resuspend sediment, thereby reducing the permanent sedimentation rate and may drive resolubilization of trace elements in the water column. The GSL has a maximum depth of about 9 m along the northwest-southeast trending fault graben in the south arm, and tapers to less than 2-3 m depth over vast areas on its west and east margins (Baskin, 2005). Thermistor strings paired with real-time meteorological data show that wind events can cause significant movement or mixing of the entire water column (Beisner et al., 2008). in flowing aI., thc aI., Wind­driven northwest­southeast aI., 2 It is necessary, therefore, to better understand the processes removing trace metals from the GSL. The trace metal of interest in this investigation is selenium (Se) due to recent concerns regarding loads to the GSL. In other terminal lakes, Se has accumulated to levels that have been associated with malformations in bird embryos and chicks, leading to decreased reproductive success (Ohlendorf and Marois, 1990). The GSL serves as a vital migratory stop for significant fractions of avian populations such as Wilson's Phalarope, Eared Grebe, and Bald Eagle (Paul, 2001). This investigation seeks to estimate the annual mass of selenium (Se) removed to sediment and verify predictions of Se volatilized to the atmosphere (Diaz et al., 2008) to determine whether sedimentation is the primary removal process. The annual Se removed by sedimentation was determined by analysis of lake sediment cores as described in Chapter 2. Verification of volatilization removal of Se was performed by direct measurement as described in Chapter 3. aI., oflake CHAPTER 2 ESTIMATING SELENIUM REMOVAL BY SEDIMENTATION Abstract The mass of selenium (Se) deposited annually to sediment in the Great Salt Lake (GSL) was estimated to determine the significance of sedimentation as a permanent removal mechanism. High Se concentrations resulting from an imbalance of removal processes with loads have resulted in reduced reproductive success of birds in other terminal lake systems. Lake sediment cores were used to delineate qualitative sedimentation regions, estimate mass accumulation rates (MAR), and determine sediment Se concentrations. Sedimentation regions were defined by comparison of isopach contours of Holocene sediment thicknesses to linear sedimentation rates of 20 short 210 226 7 137 cores. MARs were developed via analysis of Pb, Ra, Be, and Cs activity in 8 deep cores. These MARs ranged from 0.010 to 0.049 gsed/cm /yr. Contemporary sediment Se concentrations from the upper 2 cm of each deep core ranged from 0.79 to 3.12 Jig/g. Representative MAR and Se concentration were used to develop mean annual Se removal by sedimentation in the corresponding sedimentation region. The spatially integrated Se sedimentation rate was estimated to be 520 Kg/yr within a range of uncertainty between 45 and 990 Kg/yr. Comparison to annual Se loading and alternate removal processes suggests burial by sedimentation is not the primary removal process for Se from the GSL. cores. MARs were developed via analysis of 2loPb, 226Ra, 7Be, and I37CS activity in 8 deep cores. These MARs ranged from 0.010 to 0.049 gsed/cm2/yr. Contemporary sediment Se concentrations from the upper 2 cm of each deep core ranged from 0.79 to 3.121lg/g. Representative MAR and Se concentration were used to develop mean annual Se removal by sedimentation in the corresponding sedimentation region. The spatially integrated Se sedimentation rate was estimated to be 520 Kg/yr within a range of uncertainty between 45 and 990 Kg/yr. Comparison to annual Se loading and alternate removal processes suggests burial by sedimentation is not the primary removal process for Se from the GSL. 4 Introduction The Great Salt Lake (GSL) is a terminal lake located in northern Utah. To the east are the Wasatch Mountains representing the western extent of the Rocky Mountains. To the west is the Basin and Range province of the western United States characterized by extensional stress and a horst and graben landscape. The GSL resides in two distinct depressions and covers an area of about 4400 (Stokes, 1980). The south arm of the lake (defined below) contains the Carrington and East Lake Faults (Colman et al., 2002). The GSL is well known as a remnant of Pleistocene Lake Bonneville that occupied much of western Utah and stretched into Nevada and Idaho. The current geochemistry of the lake is partly a result of the evaporation of this system and partly a result of the dissolved solid load of riverine inflows (Stokes, 1980). The current GSL is a Na-Cl brine that is three to five times saltier than the ocean, but with a similar geochemical composition (Sturm, 1980). Construction of the rock-filled Union Pacific Railroad causeway in 1959 effectively divided the lake into two distinct bodies of water. The south arm, receiving the majority of freshwater inflows, has a higher water level and is less saline than the north arm (Sturm, 1980). However, the south arm is also stratified with a deep brine layer underlying the deepest portions of the lake. The deep brine layer, though transient in presence and depth, is an anoxic dense brine with high levels of sulfate reduction resulting in high sulfide concentrations (Domagalski, 1988). This deep brine layer has been thought to induce the geochemical precipitation by immobilization in sulfides of trace metals entering the lake to the point of being a "self-cleaning system" (Tayler et al., 1980). Trace metal concentrations in lake sediment were investigated in cores by Domagalski (1988). However, the resolution of these cores (both spatially and tenninallake Km2 ann aI., Stunn, ofthe ann, ann Stunn, ann aI., 5 with depth) is not high enough to interpret trace metal sedimentation rates for the whole lake since the construction of the SPRR causeway. The trace metal of interest in this investigation is selenium (Se) due to recent concerns over additional loads. High levels of Se in other terminal lakes have been associated with malformations in bird embryos and chicks (Ohlendorf and Marois, 1990). The GSL serves as a vital migratory stop for significant fractions of avian populations such as Eared Grebe, Wilson's Phalarope, and Bald Eagle. This research was performed under the context of a large Se cycling project to develop a quantitative water quality standard for the lake to protect current and future beneficial uses. This investigation seeks to estimate the contemporary mass of Se removed to sediment from analysis of lake cores for comparison to other recently measured removal processes and loads. Sediment concentrations of 24 other trace elements are also presented. The transferability of conclusions regarding Se to other trace metals is discussed. Methods Study Area The focus of this investigation is the main body of the GSL, also known as the al., south arm or Gilbert Bay (Figure 1). The vast majority of riverine inputs to the GSL flow directly or indirectly into the south arm (Tayler et aI., 1980). The south arm is defined as the lake area (exclusive of solar evaporation ponds and Farmington Bay) south of the Union Pacific railroad causeway (Baskin, 2005). Site Locations South Arm, Great Salt Lake 0 20 Figure 1. Great Salt Lake sampling locations. o 2565 ell 00 -1 iii 2267 ®~ OO.J 00 -0 o 0 OO-C OO -K o OO -B OO-A O iii o Legend 6. 2006 Deep Cores ... 2007 Deep Cores o Sha llow Cores o Se Volatilizati on Flux Measurements o 5 10 20 I I I Ki lometers OO -L iii o OO -N o 00 -0 OO -M o OO -E o OO -s o OO -Q iii OO -T o 6 7 Lake Core Collection Lake cores were taken at various sites in the south arm of the GSL. Shallow cores (~6 cm) were taken at 20 sites (yielding quantifiable sedimentation rates in 13 sites) during June 2007 across the south arm of the Great Salt Lake in order to guide the selection of locations for the deep cores collected in 2007 (described below). Core slices were collected at two intervals, 0-1 cm and 4-5 cm using a small box-coring device. After collection, samples were freeze dried and analyzed for Pb , Cs , and Be at the USGS Radioisotope Laboratory in Menlo Park, CA. A preliminary linear sedimentation 9 1 0 rate was determined in each core based on Pb decay between the two intervals using the CF-CS (constant flux-constant sedimentation) method (described below). Though these rates did not account for compaction of sediment, they were useful for determining relative differences in sedimentation rates. Deep core sediments were collected at sites 2267, 2565 and 3510 during July 2006 and at sites DD-C, DD-Q, DD-I, DD-L, and DD-R during July 2007 (Figure 1). Each of the 2007 cores was sliced into a minimum of 10 1-cm increments. At site 2267 (total water depth of 4.1 m), one gravity core of 88 cm in length was recovered. The top ten centimeters were sliced in 2-cm intervals, whereas the remainder of the core was sliced in 3-cm intervals. At site 2565 (total water depth of 8.1 m), two gravity cores (32 and 35 cm) were collected. Both were sliced in 2-cm intervals. For site 3510 (total water depth of 8.4 m), two core samples were collected. A box corer was used to collect a 12.5-cm sediment core. This device was used to avoid compaction of this shallow sediment, in order to provide the best possible determination of age as a function of depth (and sedimentation rate). This sample was sectioned in-situ in 1-cm intervals. The core ann (~6 ann ofthe oflocations 210Pb l37Cs 7Be rate was detennined in each core based on 210Pb decay between the two intervals using the CF-CS (constant flux-constant sedimentation) method (described below). Though these rates did not account for compaction of sediment, they were useful for detennining relative differences in sedimentation rates. Deep core sediments were collected at sites 2267,2565 and 3510 during July 2006 and at sites DD-C, DD-Q, DD-I, DD-L, and DD-R during July 2007 (Figure 1). Each ofthe 2007 cores was sliced into a minimum of 10 l-cm increments. At site 2267 (total water depth of 4.1 m), one gravity core of 88 cm in length was recovered. The top ten centimeters were sliced in 2-cm intervals, whereas the remainder of the core was sliced in 3-cm intervals. At site 2565 (total water depth of 8.1 m), two gravity cores (32 and 35 cm) were collected. Both were sliced in 2-cm intervals. For site 3510 (total water depth of 8.4 m), two core samples were collected. A box corer was used to collect a 12.5-cm sediment core. This device was used to avoid compaction of this shallow sediment, in order to provide the best possible detennination of age as a function of depth (and sedimentation rate). This sample was sectioned in-situ in l-cm intervals. The core 8 slices were placed into individual plastic containers and were stored on ice until transfer to a freezer. Also at site 3510, a gravity core device was used to collect a 38-cm long core, which was sliced in 2-cm intervals. The 2007 cores were collected with a gravity core device, cut into 1-cm slices and processed in a similar manner as the 2006 cores. Preparation of Core Slices for Analysis All core slices were freeze-dried under vacuum using a liquid nitrogen trap and ground using a ceramic mortar and pestle. Wet and dry weights were recorded. After grinding, the deep core slices were homogenized by mechanical mixing. Analysis of Deep Cores The homogenized slices of the 8 deep cores were divided into four fractions. One fraction was analyzed for sedimentation rate using the CF-CS method for more precise determination of sediment mass accumulation rates (MAR) in these cores at the USGS, Menlo Park, CA. In the CF-CS method, the natural logarithm of unsupported 2 1 0Pb (dpm/g) in each 1-cm increment is plotted against the cumulative dry mass (g/cm2) of 910 sediment. The decay constant for Pb divided by the slope of the linear trendline on the above plot yields the sediment MAR in gsed/cm /yr. A more thorough discussion of the analysis of radioisotopes is provided by Johnson et al. (2008). In eight cores, the second fraction was sent to the contract lab (LET Incorporated, Columbia, MO) for Se analysis by hydride generation - atomic absorption spectrometry. Core fractions were digested using proprietary digestion procedures. To reflect contemporary Se removal by sedimentation, only the top 2 cm of sediment were included l-210Pb I-cm2 ) sediment. The decay constant for 210Pb divided by the slope of the linear trendline on the above plot yields the sediment MAR in gsed/cm2/yr. A more thorough discussion of the analysis of radioisotopes is provided by Johnson et al. (2008). In eight cores, the second fraction was sent to the contract lab (LET Incorporated, Columbia, MO) for Se analysis by hydride generation - atomic absorption spectrometry. Core fractions were digested using proprietary digestion procedures. To reflect contemporary Se removal by sedimentation, only the top 2 cm of sediment were included 9 in calculating the average Se concentration for each core. The lower reporting limits for Se ranged from 0.2 to 0.4 mg/Kg. Lab results for Se concentration in the above-mentioned cores required correction for salt content. The mass of salt and additional selenium deposited on the sample from the saline pore water during the drying process was removed using the following equation to determine the Se concentration in the sediment: where [Sedry] is the concentration of selenium in the dry sample, Massw a t e r is the mass of pore water in the sample found by subtracting the dry weight from the wet weight, %Salinity is the percent salinity of the pore water, Massdry is the total dry mass of the sample, [Sesait] is the selenium concentration in the salt calculated from the percent salinity and a 0.5 Ug/L aqueous concentration, and [Sesed] is the selenium concentration in the sediment corrected for salt content. The percent salinities of the pore water in each of the cores was determined in the laboratory and ranged from 13% to 21.6% with a mean of 18%. In the three deep cores taken in 2006, a third fraction was analyzed for 24 trace elements. The fourth fraction was archived at room temperature. Analysis of trace elements was performed by ICP-MS at the University of Utah Center for Water, Ecosystems, and Climate Sciences (CWECS) laboratory. Results were corrected for salinity using the same method described for Se above. Elements included Li, Be, Al, P, [&-] Mass -(Mass i) waterx %Salinity) Massdry rS ]_ MasSwater x %Salinity x [S ] ~ e dry e salt [ ] _ Massdry Sesed - Mass(by - (Masswater x (1) Massdry Masswater MaSSdry Sesalt] J,.lg/Sesed] the sediment corrected for salt content. The percent salinities of the pore water in each of the cores was determined in the laboratory and ranged from 13% to 21.6% with a mean AI, 10 Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Sr, Mo, Ag, Cd, Sb, Ba, Tl, Pb, and U. Although Hg is another element of interest, it cannot be reliably measured in most aquatic systems using ICP-MS. Extraction of metals from freeze-dried sediment (approximately 0.5 g) was performed serially in trace metal grade nitric acid (3 mL, 15.8 N) and trace metal grade hydrochloric acid (5 mL, 12 N) using a Savillex 60-mL teflon closed reactor heated by microwave oven at 50% power for 2.5 min per reactor. The extraction solution was collected in a 50-mL centrifuge tube and made up to a volume of 50 mL with Milli-Q water. The mixture was centrifuged at 5000 rpm for 30 minutes. The supernatant was collected in a pure water-rinsed centrifuge tube while the sediments were collected in a glass Petri dish and dried at 110°C prior to weighing. For each 10th sample, a duplicate was treated via the extraction procedure and analyzed independently. Solution samples were diluted 20: 1 using 3% methanol and 0.1% HNO3 (ultra high purity) as a dilution matrix. The same matrix was used in the preparation of standards and quality control samples. Quality control was carried out using the US EPA Multi-Media, Multi- Concentration, Inorganic Analytical Service for Superfund (ILM05.3) for ICP-MS, released in February 2004 and upgraded in January 2007 (Table 1). The samples used for QA/QC (quality assurance/quality control) included an initial calibration blank (ICB), initial calibration verification (ICV), CRQL check standard (CRI), continuing calibration verification (CCV), continuing calibration blank (CCB), and interference check sample (ICS). For each 10 samples, a duplicate, spike, spike duplicated, serial dilution, CCV, and CCB were run. O.S IS.8 S SO% 2.S SO-SO SOOO 110°C 0.1 % HN03 (ultra Multi­Concentration, ILMOS.3) 11 Table 1. Quality control summary (EPA, 2007) QC Operation Frequency Instrument Calibration Initial Calibration Verification (ICV) Initial Calibration Blank (ICB) Continuing Calibration Verification (CCV) Continuing Calibration Blank (CCB) CRQL Check Standard (CRI) Interference Check Sample ( ICS) Serial Dilution for ICP Preparation Blank Laboratory Control Sample Spike Sample Post Digestion/ Distillation Spike Duplicate Sample Analysis ICP-MS Tune Method Detection Limit Determination Inter-element Corrections Linear Range Analysis Daily or each time instrument is set up. Following each instrument calibration for each wavelength or mass used. Following each instrument calibration, immediately after the Initial Calibration Verification (ICV). For each wavelength or mass used, at a frequency of 10% or every two hours of a run, whichever is more frequent, and at the beginning and end of each run. 10% or every two hours of a run, whichever is more frequent, and at the beginning and end of each run. Performed immediately after the last Continuing Calibration Verification (CCV). Every 20 analytical samples and at the beginning and end of each run, but not before the ICV. Performed before the Interference Check Sample. For ICP-AES, every 20 analytical samples and at the beginning and end of each run, immediately after the CRI. For ICP-MS, at the beginning of the run. For each matrix type or for each SDG, whichever is more frequent. For each SDG or each sample preparation and analysis procedure per batch of prepared samples. For each SDG or each sample preparation and analysis procedure per batch of prepared samples, except aqueous mercury and cyanide. For each matrix type or for each SDG, whichever is more frequent. Each time Spike Sample Recovery is outside QC limits. For each matrix type or for each SDG, whichever is more frequent. Prior to calibration. Prior to contract, annually thereafter, and after major instrument maintenance. Prior to contract, quarterly thereafter, and after major instrument adjustment. Prior to contract, and quarterly thereafter. Verification Distillation In ter-element Corrections Linear Range Analysis lCV). Perfonned CRr. frequent. frequent. limi ts. frequent. thereafter. 12 Se Removed Kg. Se = Se Cone V J f \ \Sseci J f xMAR 'sed f i „ \ x Area(ATra2)xl0 cm Kg Se {cm'yr J {Km2 jUgSeJ ' The sum of the sedimentation fluxes in each of the zones yielded the total mass of Se removed by sedimentation over the entire south arm. Sediment Traps Sediment trap pairs were deployed to capture downward sedimentation falling through the water column. The sediment traps used for sampling in the Great Salt Lake consisted of balanced pairs of detachable cylindrical acrylic sampling traps (72 mm Estimation of Se Removal by Sedimentation Annual Se removal by sedimentation was estimated from core analysis results. Holocene sediment thicknesses were estimated by David Dinter (University of Utah) and Steven Colman (USGS, Woods Hole, MA) by analysis of 30 Chirp (variable frequency) and Geopulse high-resolution seismic reflection transects (Dinter, 2007; Colman et al., 2002), as shown in the Results section. These Holocene thickness contours were plotted in ArcGIS along with the shallow core results in order to develop contours delineating qualitative zones of very high to very low contemporary sedimentation rates. Average MAR in each zone was determined by comparison of sedimentation zones to the MARs from the eight deep cores. The Se concentrations in the top 2 cm of the eight deep cores provided average contemporary sediment selenium concentrations for each zone. With Se concentration, mass accumulation rate, and area known for each of the sedimentation zones described above, the following equation was used to determine the permanent Se removal by sedimentation for each zone: of30 frequency) a1., SeRemoveJ KgSe J=seconc(j1gJXMAR( gS~d ] X Area(Km2)x1J em: K&eJ' (2) , yr gsed cm yr \.Km j1gse 13 internal diameter, 450 mm length) mounted in stainless steel holders located above their center of gravity to keep them vertical (Figure 2). The holders were attached to a stainless steel cable strung between a cement anchor and a buoy. The traps were deployed at three sites representing three distinct locations in the main body of the Great Salt Lake (Figure 1). Site 2267 was located near the mouth of the Bear River, the largest contributor of flow to the lake (70% of inflows). Sites 2565 and 3510 represent northern and southern basins in the main body of the lake. At site 2267, the top of a sediment trap pair was placed at 2.8 m below the lake surface (Figure 2), where the water depth was 4.1 m. At sites 2565 and 3510, where the water depths were 8.1 and 8.4 m, respectively, the trap pair tops were placed at two depths, approximately 27, 2006 Millipore vacuum filtration system (1.2 \xm pore size, glass micro fibre filter). The filter cake was freeze-dried, digested and analyzed for Se by ICP-MS as described above. 3.7 m and 7 m below the lake surface (Figure 3), corresponding to the shallow and deep brine layers, respectively. Sediment traps were deployed March 3, 2006 for sites 2267 and 2565, and June 27,2006 for site 3510 and were collected approximately monthly. Collected sediment in traps was corrected for salinity, measured for downward sedimentation rate and analyzed for Se concentration. After retrieving the sediment traps from the water, most of the water was drained using a peristaltic pump. The remaining water was swirled to make slurry, which was collected in 1-L polyethylene bottles and kept on ice until transfer to a refrigerator. Processing involved filtering the slurry onto a !-lm I-»• p - > • o o >-+> CO a H GO 03 bo ON - J 1.31 m Top (inlet) of trap to top of cement 0OOOOOOOOOOOC ^ P a a aq c P 23 o !-i 2 2. * 0.76 m Chain 1.8 m Cable from top of sampler assembly to float buoy % to Shallow Sediment Trap TD4.1 m Underwater Buoy -1 m under water surface .... = ~ 5 ~ ~ .... c Q.. .c... 5 .s ..... Q.. :3 ....f..:.. c Z' ~ .-.=. '-' Cement Anchor Figure 2. Schematic of sediment traps for shallow site 2267 14 Deep site Sediment Trap TD 8.1 m (2565) or 8.4 m (3510) 6 m and below - Deep Brine d S © s I 0> d S- o H a o H-Cement Figure 3. Schematic of sediment traps for deep sites 2565 and 3510 15 Cement 16 Results Mass Accumulation Rates Interpretation of the mass accumulation rates (MAR) of the 8 deep cores was performed by Chris Fuller (USGS, Menlo Park, CA). A thorough discussion on the radioisotope analysis is provided in Johnson et al. (2008). For cores in which the interpretation was more complex and multiple MAR possibilities were reported, the case was selected that incorporated the greatest number of core slices into the MAR determination. Mass accumulation rates (MAR) in the 8 deep cores ranged from 0.010 to 9 9 0.049 g/cm /yr with an average of 0.032 g/cm /yr. Cores 2267 and 2565 failed to yield 210 sufficient Pb activity for MAR determination (Table 2). Sediment Se Concentrations Salinity corrected sediment Se concentrations in the 8 deep cores are presented in |Lig/g mean of 1.73 jLig/g. Se concentrations as they apply to estimating Se removal by sedimentation are discussed below. Sediment Trace Element Concentrations Sites 3510, 2267, and 2565 were analyzed for 24 trace elements by ICP-MS at the University of Utah. Four of the 24 trace elements (Be, P, Sc, and Ag) could not be corrected for salinity because no aqueous concentration was available. Most trace element concentrations as a function of depth at site 3510 show maxima between 2-5 cm. In contrast, most trace element concentrations at the two other sites are either stable with ofthe was selected that incorporated the greatest number of core slices into the MAR determination. Mass accumulation rates (MAR) in the 8 deep cores ranged from 0.010 to 0.049 glcm2/yr with an average of 0.032 glcm2/yr. Cores 2267 and 2565 failed to yield sufficient 2lOPb activity for MAR determination (Table 2). Table 3. An example Se concentration profile and chronology is shown for site 3510 in Figure 4. Salinity concentrations for all depths ranged from 0.35 to 4.29 f.lglg with a f.lg/g. Table 2. Core mass accumulation rate results CorelD MAR g/cm2/yr) DD-Q DD-I 3510-BOX DD-L DD-R 0.027 17 CoreID (glcm2/yr) DD-C 0.036 2267-2 0.000 0.010 0.049 0.043 0.025 2565-3 0.000 Average 0.024 St. Dev. 0.019 18 Table 3. Salinity corrected deep core Se concentrations. Field ID Cone. ID Sed Se Cone. cm ng/g cm ng/3510-BOX 0-1 0.5 DD-I 0-cm 0.5 1.59 3510-BOX 1-2 2.49 DD-I 1-cm 1.5 1.80 3510-BOX 2-3 2.5 3.15 DD-I 2-3 cm 2.5 1.72 3510-BOX 3-4 3.5 DD-I 3-4 cm 3.5 2.31 3510-BOX 4-5 4.5 2.54 DD-I 4-cm 4.5 2.28 3510-BOX 5-6 5.5 1.22 DD-I 5-6 cm 5.5 2.61 3510-BOX 6-7 6.5 1.06 DD-I 6-7 cm 6.5 2.67 3510-BOX 7-8 7.5 DD-I 7-8 cm 7.5 3.59 3510-BOX 8-9 1.17 DD-I 8-cm 8.5 3.39 9.5 0.94 DD-I 9-cm 3.57 3 0-0.79 DD-L 0-1 cm 0.5 2565-3 2-4 3.0 0.90 DD-L 1-2 cm 1.5 2.26 2565-3 4-6 5.0 0.55 DD-L cm 2.63 2565-3 6-8 7.0 0.54 DD-L 3-4 cm 3.5 4.11 3 8-9.0 DD-L 4-5 cm 4.5 2.57 11.0 0.44 DD-L cm 5.5 1.09 3 12-0.67 DD-L cm 6.5 1.35 3 14-0.44 DD-L cm 0.84 2565-3 16-18 0.57 DD-L 8-9 cm 8.5 0.50 2565-3 32-34 33.0 0.77 DD-L 9-10 cm 9.5 0.49 2267-cm 1.0 1.03 DD-Q cm 0.5 3.41 2267-cm 0.45 DD-Q cm 1.5 2.82 2267-cm 0.43 DD-Q 2-3 cm 2.5 1.41 2267-2 6-8 cm 7.0 0.45 DD-Q 3-4 cm 3.5 0.91 2267-2 10-13 cm 11.5 DD-Q 4-5 cm 4.5 0.85 2 13-cm 14.5 0.67 DD-Q cm 5.5 2 16-cm 0.55 DD-Q cm 6.5 0.48 2267-2 19-22 cm 0.66 DD-Q 7-8 cm 2267-cm 26.5 0.66 DD-Q cm 0.70 2267-2 84-88 cm 86.0 0.54 DD-Q cm 0.70 DD-C 0-1 cm 0.5 3.20 DD-R 0-1 cm 0.5 1.34 DD-C cm 2.84 DD-R cm 1.5 DD-C cm 2.63 DD-R cm 2.12 DD-C 3-4 cm 2.55 DD-R 3-4 cm 3.5 2.42 DD-C cm 4.5 3.52 DD-R cm 2.75 DD-C 5-6 cm 5.5 4.29 DD-R 5-6 cm 5.5 1.74 DD-C 6-7 cm DD-R 6-7 cm 6.5 DD-C cm 3.68 DD-R cm DD-C cm 3.66 DD-R 8-9 cm DD-C 9-10 cm 9.5 3.87 DD-R 9-10 cm 9.5 0.72 10 Midpoint Sed Se Conc. Field 10 Midpoint Conc. crn !:!~/~ crn !:!g/g 2.20 00-10-1 em 1.5 00-11-2 em 00-12-em 3.35 00-13-em 351 O-00-14-5 em 00-15-em 00-16-em 1.18 00-17-em 351 O-8.5 00-18-9 em 3510-BOX 9-10 00-19-10 em 9.5 2565-30-2 1.0 OO-em 2.61 32-OO-em 34-OO-2-3 em 2.5 36-OO-em 2565-38-10 0.66 OO-em 2565-3 10-12 OO-5-6 em 2565-312-14 13.0 OO-6-7 em 2565-314-16 15.0 OO-7-8 em 7.5 17.0 OO-em OO-em 2267 -2 0-2 em OO-0-1 em 2267 -2 2-4 em 3.0 OO-1-2 em 2267 -2 4-6 em 5.0 OO-em 2267 -em OO-em 210-em 0.55 OO-em 2267-213-16 em OO-5-6 em 0.73 2267-216-19 em 17.5 OO-6-7 em 2267 -em 20.5 OO-em 7.5 0.35 2267 -2 25-28 em OO-8-9 em 8.5 2267 -em OO-9-10 em 9.5 OO-1em OO-em OO-1-2 em 1.5 OO-1-2 em 1.95 OO-2-3 em 2.5 OO-2-3 em 2.5 OO-em 3.5 OO-em OO-4-5 em OO-4-5 em 4.5 OO-em OO-em OO-em 6.5 3.68 OO-em 1.36 OO-7-8 em 7.5 OO-7-8 em 7.5 1.33 OO-8-9 em 8.5 OO-em 8.5 0.88 OO-em OO-em Figure 4. Selenium concentration profile at site 3510. 0 1 2 3 ..E.... 4 ..u... . or; 5 oIoJ Q. CLI C 6 7 8 9 10 3510-BOX Core [Se] with Depth Se Concentration (/-19/9) 0 1 2 3 4 19 5 20 depth or show maximum values near the surface. Figures 5 and 6 show concentration profiles with depth as examples for cobalt (Co) and copper (Cu) respectively. The shapes of the profiles for Co and Cu are representative of the shapes of the profiles for most other elements. Exceptions include Li, Ti, Sr. The profiles for these 3 elements at sites 2565 and 2267 are stable with depth. However, the profile for 3510 steadily decreases with depth for Li, increases dramatically between 1 and 2 cm and then stabilizes with depth for Ti, and slowly increases with depth for Sr. A table of salinity corrected results for the 20 trace elements is shown in Appendix A. Estimated Se Removal by Sedimentation Assignment of Se concentration, MAR, and area to qualitative sedimentation zones indicates that about 520 Kg of Se are removed annually by sedimentation. Results Dr. David Dinter (University of Utah) and Steven Colman (USGS, Menlo Park, CA) are shown on the map in Figure 7. The geophysical measurements used in the development of these contours are described in Colman et al. (2002). Quantifiable shallow core linear sedimentation rates ranged from 0.02 to 0.67 cm/yr. The linear sedimentation rate for core DD-I was determined to be 95 cm/yr, but is likely an outlier since it is 2 orders of magnitude greater than the remaining 12 quantifiable cores. Seven cores showed 210 negligible Pb activity and are interpreted to indicate very low sedimentation rates at these locations (DD-A, DD-D, DD-G, DD-H, DD-K, DD-O, DD-S). In general, Holocene thickness and sedimentation rates were high along the fault slightly west of the shore of western Antelope Island. East of this line, Holocene thickness decreased dramatically. West of the fault, the sedimentation rates and A. of shallow core sedimentation rates overlain on Holocene isopach contours developed by negligible 210Pb activity and are interpreted to indicate very low sedimentation rates at these locations (DD-A, DD-D, DD-G, DD-H, DD-K, DD-O, DD-S). In general, Holocene thickness and sedimentation rates were high along the fault slightly west of the shore of western Antelope Island. East of this line, Holocene thickness decreased dramatically. West of the fault, the sedimentation rates and Figure 5. Cobalt concentration profiles at sites 3510, 2565, and 2267. Note: This figure does not represent the entire length of the cores 2267 and 3510. For Co concentrations at greater depths, refer to Appendix A. Co Concentration (Jlg/g) o 2 4 6 8 o +---------~--------~--------~--------~ 2 +-----------+---~-----------~----------~ 4 +---------~--~----------~~----------~ 6 +--------+--_4~------------------------~ 8 +-------~--~----~--------------------~ -E v ~10 +---------~--~------------------------~ .c ~ 0. CI) C 12 +-- --------11------,1-----------------1 14+-------~~-4--------------------------~ 16 +------+-~----------------I 18 +-----------~--------------------------~ 20 +-------+------------~ 22 ~------------------------------------~ -+- 2267 - 2565 -+- 3510 21 figure Figure 6. Copper concentration profiles at sites 3510, 2565, and 2267. Note: This figure does not represent the entire length of the core. For Cu concentrations at greater depths, refer to Appendix A. 22 Cu Concentration (J,.lg/ g) 0 100 200 300 400 0 2 4 6 --5 180 J...:. g. 12 C 16 "U "--I 18 20 22 ~----------------------------------~ ~2267 -2565 -&-3510 3510,2565, Sediment Figure 7. Shallow core results overlain on Holocene isopach contours. Contours developed by Dinter (2007) and Colman et al. (2002). Legend Shallow Coring Results and Holocene Sedi ment Thickness South Arm, Great Salt Lake ~ DO-H. 096 0 0.12 DO-D. DD-K. 00.67 00.13 DO-A. 00.14 0 0.12 00.07 00-0. Dinter/Coleman Holocene Sediment Thickness (m) 00.02 00.07 N + 00.1 00.25 2 - 5 6-7 8-9 10 - 16 00.12 o Sedime ntation Rates (cm/yr) • Sites Where Rate Could Not Be Estimated o 5 10 20 I I I Ki lo meters 23 24 Holocene thicknesses fell more slowly and continued to decline to the western shore of the south arm of the Great Salt Lake (Dinter, 2007). The contours bounding the zones developed to reflect different sedimentation rates are shown in Figure 8. In the south basin of the south arm of the GSL, Holocene sediment thicknesses matched relatively well with shallow core results making development of the sedimentation zones straightforward. Areas with sediment thicknesses 2 m and below consistently showed 210 insufficient Pb to determine a linear sedimentation rate, indicating very low sedimentation. Areas near thicker Holocene sediment (>8 m) such as DD-C and DD-R showed the highest sedimentation rates (0.67 and 0.25 cm/yr respectively). This agreement did not, however, extend to the northwest basin of the south arm. Two shallow cores (DD-I and DD-H) and two deep cores (2565 and DD-I) fell within this basin. Holocene sediment thicknesses indicate medium to high sedimentation rates over much of the area for the past -10,000 years. However, cores DD-H and 2565 did not 210 contain sufficient Pb for a sedimentation rate determination. Though the shallow core results for core DD-I indicate that it may be an outlier, the MAR determined from the long core at this location is of similar magnitude to other cores. The discrepancy between Holocene isopach contours and shallow core results may be due to the northwest basin's proximity to the Union Pacific Railroad causeway. The sedimentation regime of this basin has likely been altered since construction. Since the shallow cores more closely represent contemporary sedimentation, Thiessen polygons were developed for the basin bounded on the east by the SW-NE trending Carrington Fault (Colman, 2002). The Thiessen polygon surrounding core DD-I was designated as a "very high" sedimentation zone. The polygons surrounding cores DD-H and 2565 were designated as insufficient 2lOPb to determine a linear sedimentation rate, indicating very low sedimentation. Areas near thicker Holocene sediment (>8 m) such as OO-C and OO-R showed the highest sedimentation rates (0.67 and 0.25 cm/yr respectively). This agreement did not, however, extend to the northwest basin of the south arm. Two shallow cores (DD-I and DO-H) and two deep cores (2565 and DO-I) fell within this basin. Holocene sediment thicknesses indicate medium to high sedimentation rates over much of the area for the past ~10,000 years. However, cores DO-H and 2565 did not contain sufficient 2IOPb for a sedimentation rate determination. Though the shallow core results for core DD-I indicate that it may be an outlier, the MAR determined from the long core at this location is of similar magnitude to other cores. The discrepancy between Holocene isopach contours and shallow core results may be due to the northwest basin's proximity to the Union Pacific Railroad causeway. The sedimentation regime of this basin has likely been altered since construction. Since the shallow cores more closely represent contemporary sedimentation, Thiessen polygons were developed for the basin bounded on the east by the SW-NE trending Carrington Fault (Colman, 2002). The Thiessen polygon surrounding core DO-I was designated as a "very high" sedimentation zone. The polygons surrounding cores OO-H and 2565 were designated as 25 Shallow Coring Results and Holocene Sediment Thickness South Arm, Great Salt Lake 0 5 10 20 I I l l I l l l I Figure 8. Qualitative sedimentation zones based on Holocene isopach contours and shallow core linear sedimentation rates. DO-A. Legend o Sedimentation Rates (em/yr) • Sites Where Rate Could Not Be Estimated ;:, 2006 Deep Cores .&. 2007 Deep Cores Sedimentation Zones _ Very High _ High Medium C] Low C] Very Low o I I I Kilometers DD-S. N + 26 "very low" sedimentation zones and grouped together. Overall, the qualitative "very low" sedimentation zone had the largest area, with areas decreasing with each increasing step in sedimentation rate. The average Se concentration in each sedimentation zone was determined by averaging the Se concentrations in the 0-2 cm interval (eg. Figure 4) for all cores falling within each zone. The average salinity corrected Se concentrations from 0-2 cm in the 8 cores are shown in Table 4. The concentrations ranged from 0.79 to 3.02 |Lig/g with an average of 2.01 jug/g. Though an MAR for cores 2565 and 2267 could not be determined, the Se concentrations in the upper 2 cm of these cores were still used to represent the most recent sedimentation. This concession was mitigated by the lower MARs associated with these zones. The Holocene thickness-based "very high SE" sedimentation zone did not contain any cores, and so was assigned a Se concentration based on that in the "high" sedimentation zone, as described below. Average mass accumulation rate (MAR) in each zone was determined by interpretation of the MAR results from the deep cores (Table 2). MARs in the "medium," "high," and "very high NW" zones were found by averaging the cores that fell within them. The MAR for the "low" sedimentation zone was calculated by averaging the two cores with sufficient Pb activity (DD-Q and 3510) with a sedimentation rate of zero for 210 core 2267 (based on insufficient Pb indicating very low sedimentation rate) - yielding an average MAR of 0.018 g/cm /yr. The "very low" sedimentation zone did not contain 210 any cores with sufficient Pb activity to estimate a MAR. Therefore, the representative MAR for this zone was estimated to be a factor of 2 below the value for the "low" sedimentation zone. The MAR for the "very high SE" sedimentation zone was estimated III cores are shown in Table 4. The concentrations ranged from 0.79 to 3.02 J..lg/g with an average of 2.01 J..lg/g. Though an MAR for cores 2565 and 2267 could not be determined, the Se concentrations in the upper 2 cm of these cores were still used to represent the most recent sedimentation. This concession was mitigated by the lower MARs associated with these zones. The Holocene thickness-based "very high SE" sedimentation zone did not contain any cores, and so was assigned a Se concentration based on that in the "high" sedimentation zone, as described below. Average mass accumulation rate (MAR) in each zone was determined by 210Pb core 2267 (based on insufficient 210Pb indicating very low sedimentation rate) - yielding an average MAR of 0.018 g/cm2/yr. The "very low" sedimentation zone did not contain any cores with sufficient 210Pb activity to estimate a MAR. Therefore, the representative MAR for this zone was estimated to be a factor of 2 below the value for the "low" sedimentation zone. The MAR for the "very high SE" sedimentation zone was estimated Table 4. Sedimentation region and mean Se concentration in cores. Se concentration represents mean of top 2-cm of core. CorelD Sed Region Sed Se Cone. 2565-3 Very Low 0.79 2267-2 Low 1.03 DD-Q Low 3.12 3510-BOX Low 2.35 DD-L Medium 2.44 DD-R Medium 1.65 DD-C High 3.02 DD-I Very High NW 1.70 Average 2.01 St. Dev. 0.86 CoreID Sed Region (Jig/g) 27 28 as 0.049 g/cm /yr, 25% higher than the "high" zone value. This value is also consistent with the representative MAR for the "very high NW" zone. Table 5 shows the Se concentration, MAR, area, and calculated mass of Se removed annually within each sedimentation zone. Results indicate that about 520 Kg of Se are permanently removed from the Great Salt Lake by sedimentation each year. Uncertainty in Se Removal by Sedimentation Uncertainty in the annual Se mass removed by sedimentation was determined by estimating uncertainty for, and propagating uncertainty through, each step in the Se removal calculation. These steps involved determining the representative sediment Se concentration, mass accumulation rate (MAR), and the area for each qualitative sedimentation zone. For the Se concentration and MAR determinations above, 2.6 in-diameter cores were used to represent the six zones with a total area of 2083 Km . The strength of this extrapolation (i.e., the greater number of cores in each zone, the stronger the confidence in the value) is incorporated into the uncertainty calculations as described below. Uncertainty in representative Se concentration for each zone was determined. Eight cores were used to describe the area of the south arm of the Great Salt Lake. In each core, the top 2 cm were sliced into one or two slices and analyzed for Se content. These concentrations were corrected for salinity as described above. Since uncertainty was not determined by the laboratory for Se concentrations in these samples, and since no replicate analyses were made, the uncertainty was estimated as two times the reporting limit (RL) for each core slice, and the uncertainties for each slice were propagated into an uncertainty in Se concentration for the core as shown, for example, for core DD-Q: glcm2/in­diameter cores were used to represent the six zones with a total area of 2083 Km2 . The was not determined by the laboratory for Se concentrations in these samples, and since no replicate analyses were made, the uncertainty was estimated as two times the reporting limit (RL) for each core slice, and the uncertainties for each slice were propagated into an uncertainty in Se concentration for the core as shown, for example, for core DD-Q: Table 5. Mean 0-2 cm Se concentration, MAR, area, and Se mass removed annually within each sedimentation zone Sed Region Area of Zone Avg. [Se] 0-2 cm Hg/g) MAR (g/cm2/yr) Mass of Se Removed (Kg/yr) Very Low 1233.2 0.79 0.009 86.06 Low 404.6 2.16 0.018 154.63 Medium 358.5 2.04 0.026 190.15 High 47.9 3.02 0.036 52.08 Very High SE 4.6 3.02 0.045 6.25 Very High NW 34.3 1.70 0.049 28.49 Total 517.65 (Km2) em em2/(J..Lg/g) 29 30 C"low" Avg ~ A/^2267 &DD-Q + ^1510 , - (4) <*w AVg = V0.82 + 0.572 + 1.132 =1.50 mg/Kg where G-iow" Avg is the uncertainty associated with averaging the Se concentration values that fall within the "low" sedimentation zone. The uncertainty in the sediment Se concentration for the entire lake is estimated as the relative standard deviation of all cores: =-?xl00% = ^ x l 0 0 % = 43% X 2.01 where RSDuke, CJ, and x are the relative standard deviation, standard deviation, and mean of the 8 cores, respectively. This is used as the background uncertainty of the entire dataset because it represents the uncertainty if the sedimentation zones described above had not been developed. These zones, though qualitative in nature, were ^DD-Q = ^o-uJ + ^ J = V0.42 + 0.42 = 0.57 mg/Kg (3) where ODD-Q is the estimated uncertainty of the average Se concentration in the top 2 cm of the core and a x . y is the uncertainty in the Se concentration in the slice of the core from depths 0 to 1 or 1 to 2 cm. The Se concentrations in the top 2 cm of cores that fall within a single sedimentation zone were averaged together to find the representative Se concentration for that zone. For example, the "low" sedimentation zone calculation of uncertainty in Se concentration is: (3) O"OD-O"x-y or 1 to 2 cm. (Y"= ~~267 + ~D-Q + ~SIO (Y"low" Avg = -J0.8 2 + 0.572 + 1.132 = 1.50 mg / Kg (4) O""[ow" RSDL ke = a xl00%= 0.86 xlOO%= 43% a X 2.01 (5) RSDLake, 0", X mean of the 8 cores, respectively. This is used as the background uncertainty of the 31 raw 43% f ?r * i ^ = 2 2 ° / 0 (6) 260 4 ^ J n / /c ore 3 cores The high area/core ratio in the "low" zone relative to that of the entire lake serves to decrease the uncertainty from the background of 43%; whereas a zone with a lower area/core ratio would have a higher RSD than 43%. This process is applied to all of the sedimentation zones, with the exception of the "very high SE" zone because no cores fell within it. The uncertainty of the "high" sedimentation zone is applied to the "very high" zone. To combine the uncertainties associated with Se concentration to those associated with extrapolation to larger areas, the RSDs are converted back to standard deviations, which are then combined as shown above. This error propagation process was repeated for each core and sedimentation zone. Uncertainty in mass accumulation rate for each zone was determined by a similar 210 method as Se concentration. However, due to the method of analysis of Pb decay (use of the slope of the trendline of the natural logarithm of unsupported 2 1 0Pb), standard developed to increase confidence in the estimation of Se removal by recognizing the spatial variation in sedimentation rates as controlled by lake bottom topography. The RSD for each zone was developed from RSD^ke by comparing the number of cores contributing information for the area. The eight cores in the 2083 Km lake yield an area/core ratio of 260 Km of lake area per core. Division of RSD^keby this value, and multiplication of the quotient by the ratio of the zone area to the number of cores in that zone yielded the RSD for each zone. The RSD for the qualitative "low" sedimentation zone (404.6 Km ) is shown below as an example: RSDLake Km2 an area/core ratio of260 Km2 oflake area per core. Division of RSDLake by this value, and multiplication of the quotient by the ratio of the zone area to the number of cores in that zone yielded the RSD for each zone. The RSD for the qualitative "low" sedimentation zone (404.6 Km2 ) is shown below as an example: RSD - 43% RSDI ake 404.6 Km2 = 2201 "low" ext 260.4 Km2 I X 3 cores 10 Icore (6) ofthe sedimentation zones, with the exception of the "very high SE" zone because no cores fell within it. The uncertainty of the "high" sedimentation zone is applied to the "very high" zone. To combine the uncertainties associated with Se concentration to those associated with extrapolation to larger areas, the RSDs are converted back to standard deviations, which are then combined as shown above. This error propagation process was repeated for each core and sedimentation zone. method as Se concentration. However, due to the method of analysis of 21oPb decay (use of the slope of the trendline of the natural logarithm of unsupported 21OPb), standard 32 deviation errors for unsupported Pb could not be propagated directly through to the final MAR value for each core. In order to determine the error associated with the MAR determination in each core, a Monte Carlo method was implemented by randomly generating an unsupported 2 1 0Pb value in each core slice using the laboratory reported unsupported 2 1 0Pb value as the mean and the 1-sigma uncertainty as the standard deviation. This was performed in Microsoft Excel using the NORMINV function paired with RAND(), which generates a random value between 0 and 1. The NORMINV function reads the RAND() value as a percentile based on the defined mean and standard deviation. For example, a mean of 2 and a standard deviation of 1 would be input as NORMINV(RAND(),2,l). If the RAND() value was 0.16, the output of the function would be 1 standard deviation below the mean (16t h percentile) yielding a value of 1. This approach was applied to each core slice to randomly generate a new unsupported 210 Pb profile, from which the slope of the linear trendline (using the SLOPE function) of the natural logarithm was determined and converted into an MAR. The process was repeated 10,000 times for each core. The uncertainty in MAR for each core was then defined by the standard deviation of the 10,000 MAR results. Figure 9 shows the convergence of the standard deviation of MAR as a function of the number of MAR values included. The standard deviations from different sized populations (10, 20, 50, 100, etc.) were determined for 10 different randomly chosen populations among the 10,000 MAR results. From this plot we observe that the range in estimated standard deviations converges to a constant after several hundred values; hence, we conclude that 10,000 repetitions is sufficient for representing the uncertainty in MAR for each core. 210Pb detennine detennination generating an unsupported 210Pb value in each core slice using the laboratory reported unsupported 210Pb value as the mean and the I-sigma uncertainty as the standard deviation. This was perfonned in Microsoft Excel using the NORMINV function paired with RANDO, which generates a random value between ° and 1. The NORMINV function reads the RANDO value as a percentile based on the defined mean and standard deviation. For example, a mean of2 and a standard deviation of 1 would be input as NORMINV(RANDO,2,1). If the RANDO value was 0.16, the output of the function would be 1 standard deviation below the mean (16th percentile) yielding a value of 1. This approach was applied to each core slice to randomly generate a new unsupported 210Pb profile, from which the slope of the linear trendline (using the SLOPE function) of the natural logarithm was detennined and converted into an MAR. The process was repeated 10,000 times for each core. The uncertainty in MAR for each core was then defined by the standard deviation of the 10,000 MAR results. Figure 9 shows the convergence of the standard deviation of MAR as a function of the number of MAR values included. The standard deviations from different sized populations (10, 20, 50, 100, etc.) were detennined for 10 different randomly chosen populations among the 10,000 MAR results. From this plot we observe that the range in estimated standard deviations converges to a constant after several hundred values; hence, we conclude that 10,000 repetitions is sufficient for representing the uncertainty in MAR for each core. 33 • • ! S i MI i $ . • • • 10 100 1000 10000 Figure 9. Standard deviations as a function of the number of randomized MAR values assessed. o E (0 o •5 « > < S 2 (0 D C ftj u.uu / 0.006 i 0.005 0.004 0.003 1 0.002 0.001 Standard Deviation in MAR Values Vs. Number of Values Assessed: Core DD-C CI 0 .007 c 0E • • IV c 0 • f I • • I I • • :6:i (I) • .!C>!! IcI:c:Ec(: 0 .004 •• • Q 0 .003 ""- IV 0 .002 "c .I.V.. (I) 0 10000 Number of MAR Values Included 34 ^ o „ e a r e a = r e a z o n e x | ! ^ - (7) A r e a i a k e where <3zom aKa is the uncertainty in the areal extent of each zone and Giake area is the uncertainty in the area of the entire lake, 1.73 Km . With uncertainties established for the sediment Se concentration, MAR, and area in each zone, the uncertainty of the mass of Se removed by sedimentation was calculated. The uncertainty was converted into a relative standard deviation for each factor in calculating the mass of Se removed: The process described for Se concentration above, propagating the uncertainty for each core to the average of the cores within a zone and then incorporating the uncertainty due to the extrapolation, was also followed for MAR uncertainty. Two cores (2565 and 2267) did not have an associated standard deviation because no MAR could be reported. In core 2565 representing the "very low" sedimentation zone, an uncertainty of 100% or 0.009 g/cm2/yr was assigned. This uncertainty was assigned in order to incorporate the interpreted MAR of the core (0 g/cm2/yr) with the assigned MAR for the zone (0.009 g/cm2/yr). Though an MAR of zero is assigned to core 2267, an uncertainty of 0.009 was assigned to this as well to be consistent with the uncertainty for 2565. The RSD of the "high" sedimentation zone (12.5%) was assigned to the "very high SE" sedimentation zone because no cores fell within this zone. Uncertainty in the areal extent of each sedimentation zone was determined. The uncertainty associated with the areal extent of the lake is 1.73 Km due to a 0.03 m stage inaccuracy in the USGS gage for lake elevation. This uncertainty was translated into uncertainties for the areal extent of each sedimentation zone by the equation: 2267) did not have an associated standard deviation because no MAR could be reported. In core 2565 representing the "very low" sedimentation zone, an uncertainty of 100% or 0.009 g/cm2/yr was assigned. This uncertainty was assigned in order to incorporate the interpreted MAR of the core (0 g/cm2/yr) with the assigned MAR for the zone (0.009 g/cm2/yr). Though an MAR of zero is assigned to core 2267, an uncertainty of 0.009 was assigned to this as well to be consistent with the uncertainty for 2565. The RSD of the "high" sedimentation zone (12.5%) was assigned to the "very high SE" sedimentation zone because no cores fell within this zone. 2 CY - A X CYiake area zone area - rea zone Area lake O"zone area O"lake area uncertainty in the area of the entire lake, 1.73 Km2 . 35 where RSD is the relative standard deviation, a is the uncertainty, and is the average value within each zone. This process was done for Se concentration, MAR, and areal extent for each sedimentation zone. The RSDs were then propagated through to the mass of Se removed in each zone and converted back into an uncertainty: RSD-™* =pStfSei + RSD2 m„+RStf area (T = RSD x Mass mass removed mass removed Se removed where cmass removed represents the uncertainty in the mass of Se removed for a particular sedimentation zone. Table 6 shows the RSD and Gmass removed values for each sedimentation zone. Since the masses of Se removed in each zone are summed to determine the mass removed for the lake, the G m a s s removed values for each scenario are propagated to define the uncertainty range: ^"total "^^"mr ^"^"mr ^"^"(^ ^0 where "mr" stands for mass removed and Gt otai mr is the uncertainty of the total mass removed. The result is a possible range between about 45 and 990 Kg Se per year, with about 520 Kg representing the mean estimate. Table 7 shows the propagation of uncertainty from the RSDmass removed for each zone to the final range of uncertainty of Se mass removed by sedimentation. RSD = ^ (8) (J RSD=­X 0" x extent for each sedimentation zone. (9) (J mass removed = RS D mass removed X MassSe removed O"mass removed sedimentation zone. Table 6 shows the RSD and O"mass removed values for each O"mass (Jtotal mr = o:.very low" mT + o:.low" mr + o:.medium" mT + o:.high" mr + o:.very high" mr (10) O"total RSDmass removed mass removed by sedimentation. 36 Table 6. Total zone relative standard deviation determination. Total zone RSD determined from RSD of Se concentration, area, and MAR. Sed Region RSD [Se] RSD Area RSD MAR Total Zone RSD Very Low 2.27 0.00083 3.89 4.51 Low 0.73 0.00083 0.70 Medium 0.49 0.00083 1.03 1.14 High 0.20 0.00083 0.19 0.28 Very High SE 0.20 0.00083 0.19 0.28 Very High NW 0.34 0.00083 0.17 0.38 Table 7. Mean and range of estimated annual Se removal by sedimentation. Range based on estimated removal of 517.65 Kgse/yr. Total Zone RSD Mass of Se Total Zone Sed Region TotRalS DZ one Removed Uncertainty (Kg/yr) Very Low 4.51 86.06 387.69 Low 1.01 154.63 156.38 Medium 1.14 190.15 217.55 High 0.28 52.08 14.57 Very High SE 0.28 6.25 1.75 Very High NW 0.38 28.49 10.82 Total 517.65 471.61 Range of 46.05 989.26 Se) 1.01 HighNW KgsJ ofSe (Kg/yr) Removal 37 Sediment Traps Salinity corrected sediment trap accumulation results showed significant spatial trends (Figures 10-12 and Tables 8-12). The mass of sediment that accumulated in the traps represents the downward sedimentation flux at that location over the period of deployment. The sediment trap at shallow site 2267 yielded an average downward sediment flux of 2.95 g/cm2/yr for the period 03/23/06 to 06/26/07, which is an order of magnitude higher than any of the other average sediment fluxes measured during that period. The next-highest apparent sedimentation rates occurred at the two deep sites (2565 & 3510, Tables 9 and 11), which were 0.53 and 0.35 g/cm2/yr, respectively. The shallow sediment traps at sites 2565 and 3510 (Tables 10 and 12) yielded very low downward sedimentation rates (0.035 and 0.068 g/cm2/yr) that were approximately an order of magnitude below those of the corresponding deep traps. The high sedimentation rates at shallow site 2267 correspond to its location in a relatively narrow channel between the Promontory Point and Fremont Island near the outlet of the Bear River. The observed peak sedimentation rate in spring corresponds to peak discharge from the Bear River. The high sedimentation rates in the deep traps relative to the shallow traps at sites 2565 and 3510 likely reflect resuspension of sediment from the lake bottom, since it is unlikely that sediment was generated at intermediate depths. Had the material in the deep traps originated from shallower water, it would have also been collected in the shallow traps. This observation indicates significant resuspension of lake-bottom sediment. The topic of resuspension will be further described below. deployment. sitc of2.95 cm2/traps. This observation indicates significant resuspension of lake-bottom sediment. The topic of resuspension will be further described below. Figure 10. Sedimentation and Se sedimentation fluxes at shallow site 2267. The Se flux values plotted were multiplied by 106. To obtain the actual values, multiply by 10"6. Period of measurement was February 2006 to July 2007. 38 ~S entation flux ---=-- Se e~inherltation u -- 5 -I. - 1- -- -r- - -1 I -4 ~ ->- t- 'J1 N ~ E 1-- - = -tO)' l 4 - j :~:: ---- -, ~ >< ~ -~.:: 3 - 3 e:~ c ,,-.. t- 0 ~ 1- ~ (') ..... a c 2 - 2 N -- Q) '< E -.,- :.0 Q) (J) 1 ,./ 0 Feb A 12 13 1 PJUr;h,.12 ugdJt 60 2 ~, Dqt.e ec F9e b A7 prBJu n 7 4 3 <::>0~ ~ 106 . 10-6 . Period of measurement was February 2006 to July 2007. 39 Figure 11. Sedimentation and Se sedimentation fluxes at deep site 2565. The Se flux 7 - y values plotted were multiplied by 10 . To obtain the actual values, multiply by 10 . Period of measurement was February 2006 to July 2007. --Sedimentation flux - deep --Sedim lux - shallow 4 I Se s[dime nt~tioln flux - d\3ep 1 8 -- ---+- ~e Sl'dr ntatioln ~ux 1- s~alllbw I ~ ->- \ I I _ ~ _ -- - -[- - - - 4 I N I - \ E - -- - I 1 -1 1 <..) 1- - -6 rJJ 3 I I I 1 - ~ .O.-l = >< I ~ 1 = ::::l -j r ~ ~ I .,- ~ c I 4 .... 0 0 2 1-- ~ ! ·1 .!... :g I ,-., IJQ ~c -,,. -n- (l) - t 8 -~ r 2 N -- -u 1 I '<.,! (l) '-' (/) F eb,214 15 .- - . P'jun. 14 AU9dc?b12 ec 11 D Feb 9 Cite Apr 10 Jun 9 7 8 11 . values plotted were multiplied by 107. To obtain the actual values, multiply by 10=7. Period of measurement was February 2006 to July 2007. Figure 12. Sedimentation and Se sedimentation fluxes at deep site 3510. The Se flux values plotted were multiplied by 108. To obtain the actual values, multiply by 10"8. Period of measurement was February 2006 to July 2007. - Sedimentation flux - deep --'>--- 4 8 N E -() 3 CJ) I -->< I - -, :;:) ;:;::: L 1--"- [ 1-- r c: 0 2 i ~ I .- I c: __ l (1) I E 1-- 1 -" I ""0 , . I I . -rI - 4 2 (1) Cf) 8 35 10. 108 . 10-8 . 40 Table 8. Results of site 2267 shallow sediment trap analyses Average month Days accum. Average sediment Downward flux (g/cm2/year) [Se] (mg/Kg) Se downward flux (gSe/cm /yr) 6-Apr 64 18.22 2.55 0.27 6.76E-07 6-Jun 24 8.81 3.29 1.54 5.07E-06 6-Jul 39 18.21 4.19 0.31 1.30E-06 6-Jul 32 9.96 2.79 0.2 5.58E-07 6-Sep 64 24.76 3.47 0.33 1.13E-06 6-Nov 36 8.49 2.12 0.2 4.23E-07 7-Jan 103 8.82 0.77 1.16 8.94E-07 7-Apr 37 21.65 5.25 n/a n/a 7-Jun 34 8.13 2.14 n/a n/a Average 2.95 0.57 1.44E-06 Accumulative 433 127.06 2.63 n/a - not available Table 9. Results of site 2565 deep sediment trap analyses Average Se sediment Downward downward Average Days weight flux [Se] flux month acum. (g/cm2/year) (mg/Kg) (gSe/cm2/yr) 6-Apr 64 14.29 2 0.02 3.00E-08 6-Jun 24 0.86 0.32 1.7 5.48E-07 6-Jul 39 0 0 n/a n/a 6-Aug 45 0.55 0.11 1.67 1.83E-07 6-Oct 51 1.26 0.22 0.23 5.09E-08 7-Jan 139 1.24 0.08 0.51 4.04E-08 7-Apr 37 1.79 0.43 0.15 6.50E-08 7-Jun 42 5.02 1.07 Average 0.53 0.71 1.53E-07 Accumulative 441 25.01 0.51 n/a - not available 41 weight (g) cm2/cm2/1. 13E-(g) year) cm2/0ct 42 Table 10. Results of site 2565 shallow sediment trap analyses Average Se sediment Downward downward Average Days weight flux [Se] flux month acum. (g) (g/cm2/year) (mg/Kg) (gSe/cm2/yr) 6-Apr 64 0 0 n/a n/a 6-Jun 24 0 0 n/a n/a 6-Jul 39 0 0 n/a n/a 6-Aug 45 0.169 0.034 1.279 4.30E-08 6-Oct 51 0 0 n/a n/a 7-Jan 139 0.853 0.055 0.558 3.07E-08 7-Apr 37 0.512 0.124 0.445 5.52E-08 7-Jun 42 0.301 0.064 Average 0.035 0.761 4.30E-08 Accumulative 441 1.834 0.037 n/a - not available Table 11. Results of site 3510 deep sediment trap analyses Average month Days acum. Average sediment Downward flux (g/cm2/year) [Se] (mg/Kg) Se downward flux (gSe/cm2/yr) 6-Jul 30 3.61 1.08 0.01 1.08E-08 6-Aug 47 0.08 0.02 0.87 1.41E-08 6-Oct 52 0.85 0.15 0.18 2.55E-08 6-Nov 34 0 0 n/a n/a 7-Jan 101 1.09 0.1 0.7 6.82E-08 7-Jun 28 2.35 0.75 0.1 7.52E-08 Average 0.35 0.31 3.88E-08 Accumulative 292 7.98 0.25 n/a - not available (g) cm2/yr) 0ct weight (g) cm2/cm2/~r) 0ct Table 12. Results of site 3510 shallow sediment trap analyses Average Se sediment Downward downward Average Days weight flux [Se] flux month acum. (g) (g/ cm2/year) (mg/Kg) (gSe/cm2/yr) 6-Jul 30 0 0 n/a n/a 6-Aug 47 0.1 0.02 1.44 2.85E-08 6-Oct 52 0.35 0.061 0.2 1.19E-08 6-Nov 34 0 0 n/a n/a 7-Jan 101 0.82 0.072 0.76 5.50E-08 7-Jun 28 0.79 0.252 0.01 < 2.52E-09 Average 0.068 0.4 3.18E-08 Accumulative 292 2.06 0.063 n/a - not available 43 flux flux cm2 Iyear) cm2/y"r) 0ct < 0.01 44 In terms of temporal variation, all sites showed higher sedimentation rates in spring and early summer relative to late summer and fall (Figures 10-12 and Table 8-12). The average Se downward fluxes mirror the spatial trends in downward sediment fluxes (Figures 10-12 and Table 8-12), where the average downward Se flux at shallow site 2267 (1.44x10" g Se/cm /yr) was one to two orders of magnitude larger than those at the deep sites (2565 & 3510, Tables 9 and 11), which were 1.53xl0"7 and 3.88xlO"8 g Se/cm /yr, respectively. The downward Se flux obtained at the shallow sediment traps at sites 3510 and 2565 yielded 3.18xl0"8 g Se/cm2/yr and 4.30xl0"8 g Se/cm2/yr (Tables 12 and 10). Regarding temporal variations, peak downward Se fluxes did not correspond to peak sedimentation fluxes (Figures 10-12), and did not show an apparent correspondence to season. However, the limited data would not be expected to yield a clear trend. Collected sediment included mineral particles and organic material (e.g., phyto-and zoo-plankton, and brine shrimp). Based on visual inspection, mineral particles dominated the matrix at site 2267, whereas accumulated sediments at the other sites appeared to have mostly organic material. A notable exception occurred at site 2565 in April 2006 when the matrix was dominated by mineral particles and the sedimentation flux was relatively high. The downward sedimentation rates will be compared to permenant sedimentation rates below. oftemporal t1ux l.44x10-6 cm2/1.53x10-7 3.88x10-8 cm2/3.18xlO-8 4.30xl0-8 phyto­and 45 Discussion Comparison of Removal to Loading of Se Comparison of the estimation of Se removal by sedimentation with estimation of loading of Se to the system indicates that removal by sedimentation accounts for less than half of the Se mass entering the main body of the GSL. Annual loading is estimated as 1480 Kg Se/yr (Naftz et al., 2007). This load is well above the mean estimated Se removal by sedimentation of 520 Kg and even greater than the upper estimate based on uncertainty (990 Kg Se/yr). Estimations and direct measurements of Se removal to the atmosphere indicate that volatilization, not sedimentation, is the main mechanism of Se removal from GSL (Diaz et al., 2008). The mean estimate of annual Se removal by volatilization was 2108 Kg/yr. Sediment Traps Comparison of sediment trap results to permanent sedimentation rates indicates that significant sediment resuspension and redistribution is occurring in the South Arm of the GSL. A representative downward sedimentation flux from the shallow sediment traps at sites 2565 and 3510 can be considered to be representative of the main body of the Great Salt Lake. Representative sedimentation fluxes could not be obtained from site 2267 due to its proximity to the Bear River. Nor could such a flux be obtained from the deep sediment traps at sites 2565 and 3510, due to the influence of resuspension. The average sedimentation rate for these two shallow sediment traps was 0.016 g/cm /yr. This value is lower than the permanent sedimentation rate from the core at site 3510 (0.043 g/cm /yr), indicating that the permanent sedimentation rate does not reflect through-fall ofthe aI., aI., cm2/cm2/• 46 from the surface. This discrepancy indicates that resuspension and lateral transport of newly deposited sediment to permanent deposition zones is significant and in agreement with the Be results (Johnson et al. 2008). Regarding downward Se sedimentation rate, the single significant value for the shallow sediment trap at site 3510 was 1.19xl0"8 g Se/cm2/yr. This value is smaller than 8 2 the permanent Se sedimentation rate from the core at site 3510 (4.2x10" g Se/cm /yr). Based on relative overall sedimentation rates, one might have expected the downward Se sedimentation rate to exceed the permanent Se sedimentation rate at site 3510 (reflecting resuspension). However, lateral redistribution of Se is expected to occur as a result of resuspension in the deep brine layer. Recall that Se accumulation in the deep traps at site 8 2 3510 was 1.4x10" g Se/cm /yr, which matches the order of the permanent Se sedimentation rate (4.2x10" g Se/cm /yr). Trace Elements The transferability of our observations regarding Se to other trace elements cannot be assessed directly since the loads of other trace elements to the GSL were not measured. However, some inferences can be made due to the relationship between Se and other trace elements. First, the different trends between cores detailed in the Results section likely occur due to differences in sedimentation rate. Trace element concentrations at sites 2267 and 2565 are largely stable at depth, but some show significant increases near the surface. Both of these sites have very low sedimentation rates as shown above. If increased loadings due to recent mining and urban development resulted in increased sedimentation of a trace element, this would be reflected only in the upper slice(s) of the cores. 7Be 1.19xlO-8 the pennanent Se sedimentation rate from the core at site 3510 (4.2xlO-g Se/cm2/yr). Based on relative overall sedimentation rates, one might have expected the downward Se sedimentation rate to exceed the permanent Se sedimentation rate at site 3510 (reflecting resuspension). However, lateral redistribution of Se is expected to occur as a result of resuspension in the deep brine layer. Recall that Se accumulation in the deep traps at site 3510 was 1.4xl0-8 g Se/cm2/yr, which matches the order of the permanent Se sedimentation rate (4.2xl0-8 g Se/cm2/yr). Trace Elements The transferability of our observations regarding Se to other trace elements cannot be assessed directly since the loads of other trace elements to the GSL were not measured. However, some inferences can be made due to the relationship between Se and other trace elements. First, the different trends between cores detailed in the Results section likely occur due to differences in sedimentation rate. Trace element concentrations at sites 2267 and 2565 are largely stable at depth, but some show significant increases near the surface. Both of these sites have very low sedimentation rates as shown above. If increased loadings due to recent mining and urban development resulted in increased sedimentation of a trace element, this would be reflected only in the upper slice(s) of the cores. Alternatively, the higher sedimentation rate at 3510 may more clearly reveal recent changes in trace element removal to sediment. As mentioned before, most elements analyzed at site 3510 show a peak in concentration between 2 and 5 cm depth. Though only one deep core with a quantifiable sedimentation rate was analyzed for the 24 trace elements (3510), the correspondence between the Se concentration profile and the profiles of the other trace elements is compelling. It has been shown that diagenetic changes can influence the concentration-depth profiles of trace elements such that historical trends may not be accurately recorded (Callendar, 2000). However, one would not expect this effect to be consistent among different trace elements within a core. The consistency with which the trace element concentration profiles mirror that of Se suggests that the shape of the profiles may be determined by changes in loading as opposed to diagenetic processes. To further investigate this possibility, the Se concentration profiles were converted to chronologies based on the MARs determined by Chris Fuller (USGS, Menlo Park, CA). These chronologies are shown in Figure 13 and Figure 14 for the six deep cores with quantifiable MARs (excludes 2267 and 2565). Results show that sediment Se concentrations deposited after 1950 are consistently higher than the concentrations of Se in the sediment prior to 1950. Also, the concentration of Se decreases in the sediment during the 1980s and 1990s after the observed peak in four of the six cores. If trace elements consistently behave in a similar manner to Se in the sediment in the GSL, as they do in core 3510, then Figures 13 and 14 indicate that many different trace elements were being removed to sediment at higher rates after 1950. 47 48 0 1 2 3 4 5 Se Concentration (mg/kg) BOX-3510 DD-C DD-I Figure 13. Se concentration chronology based on MAR in cores at sites 3510, DD-C, and DD-I. Sediment [Se] Chronology 2025 2010 1995 1980 ,....., ..5.... 1965 ..c..:, ~ 1950 c 1935 1920 1905 1890 ~BOX-3510 -DD-C -...-DD-J. E u o. Q 0 1 2 3 4 5 Se Concentration (mg/kg) Figure 14. and DD-R. Se concentration chronology based on MAR in cores at sites DD-L, DD-Q, Sediment [Se] Chronology 2025 2010 1995 1980 1965 ~ 1950 - 5 1935 "'-" ...c., 1920 ~ 1905 Q 1890 1875 1860 1845 1830 1815 -+- DD- L _ DD-Q --.- DD- R 49 50 The universality of increased Se concentrations (and likely other trace elements) in sediment deposited after 1950 among the cores examined suggests increases in loads during that period. The extent to which increases in trace element sedimentation balance loads, however, cannot be examined without knowing the magnitude of the loads. More research is also required to determine the significance of sedimentation relative to other removal mechanisms for these trace elements. Conclusion Though significant geochemical processing and removal of Se to sediment occurs, sedimentation alone does not balance incoming loads. Mean estimated Se removal by sedimentation is about 520 Kg/yr with a range of uncertainty between approximately 45 and 990 Kg/yr - well below the 1480 Kg/yr estimate of loading to the system (Naftz et al., 2007). This investigation shows that sedimentation is a major, but not the primary removal mechanism for Se from the Great Salt Lake. More research is required to determine the transferability of this conclusion to other trace metals. ofloading aI., CHAPTER 3 VERIFYING SELENIUM REMOVAL BY VOLATILIZATION Abstract Direct measurements were made of volatilized selenium (Se) from the surface of the Great Salt Lake, UT (GSL) to verify the applicability of a predictive model for Se flux based on measured dissolved Se concentrations, wind speeds, and water temperature. An emission isolation flux chamber (St. Croix Sensory, Inc.) floating on the water surface circumscribed a capture area. Helium released into the chamber swept volatilized Se to a temperature controlled cryogenic trap (-170°C). An intermediate water trap in an acetone/dry ice slush (-20°C) removed water vapor before reaching the cryogenic trap. The captured mass of volatilized Se over the multiple-hour samples was determined by ICP-MS. Measurements of surface water temperature, wind velocity, and aqueous volatile Se concentration were made concurrently with direct flux measurements for input into the predictive model. Direct flux measurements under controlled laboratory conditions in which Se concentration varied and was independently verified suggest that 10% of actual flux was captured by the direct measurement. After correction for background and 10% measurement inefficiency, measured fluxes approximated, but were generally higher than predicted fluxes. The specific cause of this discrepancy is unclear, but the correspondence is strong enough with the limited number of direct measurements CSt. acetone/dry ice slush (-20°C) removed water vapor before reaching the cryogenic trap. The captured mass of volatilized Se over the multiple-hour samples was determined by ICP-MS. Measurements of surface water temperature, wind velocity, and aqueous volatile Se concentration were made concurrently with direct flux measurements for input into the predictive model. Direct flux measurements under controlled laboratory conditions in which Se concentration varied and was independently verified suggest that 10% of actual flux was captured by the direct measurement. After correction for background and 10% measurement inefficiency, measured fluxes approximated, but were generally higher than predicted fluxes. The specific cause of this discrepancy is unclear, but the correspondence is strong enough with the limited number of direct measurements 52 that a correction of the predicted flux is not warranted. This investigation shows that significant volatilization of Se is occurring on the GSL and that prediction of volatile Se flux using measured volatile Se concentrations and meteorological parameters represents an appropriate means for estimating annual Se removal by volatilization on the Great Salt Lake. Introduction In the previous chapter it was shown that mean estimates of Se removal from the GSL by permanent sedimentation account for less than half of the estimated loads. In ocean environments, another removal process, volatilization, has been shown to be significant removal mechanism (Amouroux, 2001). Estimates of volatilizaton can be made based on Fick's law of air-water exchange rates (Amouroux and Donard, 1997; Schwarzenback et al., 2003). This model incorporates the concentration gradient and the air-water transfer velocity, kw . The air-water transfer velocity must be adjusted, however, to the unique, hypersaline environment of the GSL. Volatile compounds of Se can form from inorganic dissolved Se such as selenite or selenate through biomethylation by phytoplankton (Amouroux and Donard, 1996). Diaz et al. (2008) at the University of Utah applied a predictive model to estimate annual Se removal by volatilization from the GSL. In order to assess whether volatilization of Se is occurring on the GSL and the validity of the predictive model, direct measurements of volatilized Se were made from the surface of the lake using a floating emission isolation flux chamber and temperature controlled cryogenic trapping system. ofvolatilizaton aI., kw. aI. 53 As with the investigation of Se removal by sedimentation in Chapter 2, this research was performed in the context of a larger Se cycling project to assist in the development of a quantitative water quality standard for Se on the GSL. Methods Study Area The focus of this investigation is the main body of the GSL, also known as the south arm or Gilbert Bay (Figure 1). The vast majority of riverine inputs to the GSL flow directly or indirectly into the south arm (Tayler et al., 1980). Concurrent GSL investigations of Se loads and other removal mechanisms also focus on the main body of the GSL. The south arm is defined as the lake area (exclusive of solar evaporation ponds and Farmington Bay) south of the Union Pacific railroad causeway (Baskin, 2005). Measuring Volatile Se Concentration kPhi/Ph2, cryogenic trap in liquid nitrogen (-196°C) collected volatile compounds from the flowing vapor for analysis by ICP-MS. aI., The volatile Se flux predictive model developed by Diaz et al. (2008) is defined by the mass transfer velocity, kphllph2, and the concentration gradient between the water and vapor phases. To determine a gradient, concentration of volatile Se in the water phase was measured by a purge and cryogenic trapping system following concepts by Amouroux and Donard (1996). Helium was swept through 7 liters of GSL water in a reactor (modified desiccator) to purge volatile Se into the vapor phase. An initial dry ice/ethanol water trap (-55°C) kept water vapor from entering the cryogenic trap. The 54 Calibration tests of the purge and cryogenic trapping system using spiked dimethyl selenide (DMSe) showed consistent 25% recovery of volatile Se from water. DMSe is assumed to be the dominant volatile Se species (Amouroux and Donard, 1996). Further details of this trapping system as described in Diaz et al. (2008) are found in Appendix B. Volatile Se Flux Predictive Model The water transfer velocity, kx , was estimated based on the wind velocity and Schmidt number. The Schmidt number (dimensionless ratio of viscosity to diffusivity) determination is based on surface water temperature and was corrected for salinity. The wind velocity was measured as described below. Further details of the Se flux predictive model as described in Diaz et al. (2008) are found in Appendix C. Inputs into the predictive model were measured concurrently with direct measurements. Wind velocity was measured at 3 m height above the water surface with a Kestrel 1000 Wind Meter and averaged over the course of the sample. Wind measurements were projected to a height of 10 m by the method described by Wind Energy Department of Risoe National Laboratory and Det Norske Veritas (2001) for use in flux prediction calculations. Water temperature at the surface (0.2 - 0.5 m depth) was measured using a Hydrolab Troll 9000 (In-Situ Inc., Fort Collins, CO). Volatile Se concentration was measured as described above and in Appendix B. Direct Se Flux Measurement The direct flux measurement system was developed and implemented to contain and capture volatilized Se from the lake surface. Direct measurements of volatilization of Appendix B. kx, diffusivity) 10m ofRisoe measured using a Hydrolab Troll 9000 (In-Situ Inc., Fort Collins, CO). Volatile Se concentration was measured as described above and in Appendix B. Se were taken at two primary locations (3510 and 2267) and one secondary location (2565) in the south arm of the Great Salt Lake (Figure 1). The flux measurements were taken concurrently with characterizations of the parameters used in estimating volatile Se flux: surface water temperature, wind velocity, and volatile Se concentration, in order to assess the accuracy of the predictive model. An emission isolation flux chamber (St. Croix Sensory, Inc.) was used to collect volatilized Se from the surface of the lake (Figure 15). The cylindrical bottom of the L/min al., L/min of the volatile Se was cryo-focused onto glass wool in a finger-trap held at -170°C by liquid nitrogen and a Watlow PID temperature controller connected to a temperature sensor (PT-103-AM Platinum RTD, Lakeshore Cryotronics, Inc.) and cartridge heater (3039- 55 ofthe stainless steel chamber circumscribed a capture area for volatile compounds. Helium gas was released from a compressed helium tank and swept through the chamber (while floating on the lake surface) to drive volatile gases coming from the lake into a cryogenic trap. The sweep rate was set to approximately 3 Umin to prevent accumulation of volatilized Se (and other gases) within the chamber. A constant sweep rate was used in lieu of a variable rate matching environmental conditions because studies have shown that high sweep rates can induce convection in the water column and subsequently bias flux results high (Card et aI., 2002). A sweep rate of 3 Umin corresponds to approximately 1 chamber volume being swept every 6 min and is consistent with the manufacturer's recommendations. The gas mixture in the chamber was then pumped (Universal 44XR Single Pump, SKC West Inc.) at an equivalent rate through Teflon tubing to a glass finger-trap in an acetone/dry ice slush (-20°C) to remove any water vapor. Downstream ofthe water trap, nitrogen and a Watlow PID temperature controller connected to a temperature sensor (PT-I03-AM Platinum RTD, Lakeshore Cryotronics, Inc.) and cartridge heater (3039- 56 Figure 15. St Croix Sensory, Inc. Emission Isolation Flux Chamber during sample collection on calm day. 57 002, Cryogenic Control Systems, Inc.). Figure 16 depicts the apparatus used to hold the glass finger-trap in the liquid nitrogen. Designed with the assistance of Dr. Kip Solomon (University of Utah) and Erwin McPherson (University of Utah), the device was placed in a Dewar flask filled with liquid nitrogen. of 20% CC-ICP- MS facility. The resulting measured concentration was then converted to a mass of Se and divided by the area under the flux chamber (0.13 m ) and the period of sampling in order to yield a flux rate. Verifying Proper Measurement System Performance To ensure that the sampling system was operating properly, tests were performed to quantify the background level of Se, examine response of the system to changes in volatilization rate, and verify reproducibility of measurements. Three flux samples were taken in the laboratory by placing the chamber over a nitric acid-washed pan filled with pure water (Milli-Q) to determine the background "flux" that is measured in a pure The "heat" of the liquid nitrogen was conducted through the brass rod to the copper block and tube surrounding the finger-trap. The length of brass rod necessary for optimal temperature controller performance was determined experimentally to be 1 cm, at which point the cartridge heater embedded in the copper block was activated approximately 25% of the time. The stainless steel shield prevented any direct contact between the liquid nitrogen and the copper block and tube. After a substantial sampling time (typically between 1.5 and 3 hours), the sample in the cryo-trap was acidified with 5 mL of20% nitric acid to stabilize volatile Se compounds as oxidized aqueous species. The sealed trap was then digested in a water bath at 75°C for 3 hours and analyzed for Se by ICP-MS at the University of Utah CC­ICP- m2 ) Figure 16. Diagram of temperature-controlled cryo-focusing system for collection of volatilized selenium from GSL. ,. --> Flow Out ,'~ Glass Trap I~ J Flow In - - - -> I r- I I ! I I I I I I' I I I I ! I ( ,.. -t~sy Wool f- Copp0er Block ~~Jl Heater ~ ''('""(! as U 58 59 sample. All background samples were low, with a mean of 1.60 ng/m h and a maximum of 2.67 ng/To test response of the system to changes in flux rate, two more samples were taken at Saltair marina for 30 min each. During the first sampling, a diffuser hanging 1 m below the surface bubbled helium through the water column into the flux chamber to produce a high flux rate. For the second sample the diffuser was turned off, producing a low flux rate. Analyses yielded an order of magnitude higher flux rate for the first sample relative to the second, indicating that the system responded appropriately. Reproducibility was demonstrated by two 2-hour samples taken during the same day at site 2267. The results showed similar Se flux rates with a slightly higher flux rate corresponding to the sample that was taken under conditions of increased surface chop. The effect on measured flux of sweep rate, sweep gas composition (helium versus nitrogen), and concentration of dissolved volatile Se was also investigated. In the laboratory, a stainless steel basin was filled with 50 L of Great Salt Lake water and spiked with various masses of DMeSe to yield concentrations from about 2 to 27 ng/L. Seven liters of this water were analyzed for volatile Se concentration using the purge and cryo-trap system described above. Other input variables were held constant with a steady water temperature and no wind. The flux of volatile Se from the solution remaining in the stainless steel basin was measured using the system described above. Sweep rates were 2 and 3 L/min for helium, and were 2,3, and 6 L/min for nitrogen. All five flux measurements were higher than those observed on the Great Salt Lake, consistent with the higher dissolved volatile Se concentration in the chamber relative to the lake (Table 13). The measured fluxes were also highly reproducible regardless of whether nitrogen m2h m2h. ofthe ofDMeSe Llmin Llmin l3). Table 13. Results of measured volatile Se fluxes under controlled laboratory conditions with variable sweep rate and sweep gas composition. Sample ID Sweep Gas Sweep Rate L/min Wind m/s Water Temp. C Vol [Se] ng/L Measured Flux ng/GT1 N2 2 0 19 2.24 5.72 GT2 N2 3 0 19 21.50 32.26 GT3 N2 6 0 19 17.45 24.65 GT4 3 0 19 27.36 43.11 GT5 2 0 19 21.90 29.51 GT7 N2 2 0 19 21.90 31.80 IX 3 0 19 0 2.67 2X 3 0 19 0 1.25 3X 3 0 19 0 0.87 60 mls °c nglL m2h GTl 21 .50 He He He He He 61 or helium gas was used, and the values did not change as a result of changing sweep rate. The influence of concentration is discussed in the results section. A recovery test was performed in the laboratory in order to quantify the response of the direct measurement system to the introduction of a known mass of Se. A DMSe spike solution was prepared to a concentration of approximately 125 jLig/L. Drops of the spike solution were placed on a glass Petri dish and immediately set inside a large cylindrical Pyrex basin. The flux chamber was then fitted tightly over the basin and a 1- hour direct measurement was performed as the drops of spike solution evaporated into the flux chamber headspace. No outside heat was applied to speed evaporation. Three recovery tests were performed with 50, 100, and 150 jaL of spike solution. The mass of spiked Se was verified independently by analysis of equivalent volumes of spike solution by ICP-MS using the digestion procedure described above. The results of the flux recovery tests are discussed below. Results and Discussion Sensitivity of Measured Fluxes to Input Variables Measured fluxes ranged from 2 to 20 ng/m h among the seven measurements m/s, temperatures of 12 to 28°C, and volatile Se concentrations of 0.04 to 4.6 ng/L. The results of measured volatile Se fluxes from Great Salt Lake are shown in Table 14 along with the corresponding predicted flux rates (Estuarine model) and input variables (wind velocity, surface water temperature, surface volatile Se concentration). The measured volatile Se fluxes were highly sensitive to near-surface volatile Se concentrations. For example, despite similar wind and wave conditions, samples 1C and ofa ofSe. /-lg//-lL m2h made on the Great Salt Lake, with average wind velocities of 1 to 4 mis, water ofO.04 1 C Table 14. Measured and predicted volatile Se fluxes with corresponding environmental parameters. Sample ID Site Date Avg. Wind Vel. Surface Temp. Vol Se Cone. Measured Flux Estuarine Pred. Flux m/s °C ng/L ng/m2h ng/m2h IB 3510 6/1/07 2.63 20.55 0.21 11.12 6.20 1C 3510 6/27/07 3.74 24.71 0.04 2.08 2.00 2C 2267 7/2/07 1.25 25.53 1.82 20.13 39.64 3C 2267 7/26/07 1.34 27.09 4.59 9.38 105.89 4C 2565 7/27/07 1.86 27.67 0.37 3.23 10.12 IE 2267 9/27/07 3.43 12.00 0.62 7.85 19.75 2E 2267 9/27/07 1.58 12.00 0.33 3.30 5.68 62 °c m2h m2h 1B 1 /lC 63 2C yielded volatile Se fluxes of 2.08 and 20.13 ng/corresponding to measured near-surface volatile Se concentrations of 0.04 and 1.82 ng/L, respectively. Measured volatile Se fluxes did not change appreciably with changes in wind speed under the relatively calm conditions examined. For example, in two samples taken on 5/24/2007 to show reproducibility, measured flux rates were 10.7 and 8.0 ng/mii for average wind velocities at 10 m above the water surface of 5.1 and 6.7 m/s respectively. Though counterintuitive, the surface roughness during the first sample was significantly higher than the second resulting in the slightly higher flux rate for the first (assumes constant volatile Se concentration and water temperature). This observation is consistent with the fact that surface matrix effects, rather than wind speed, dominate liquid-to-atmosphere fluxes on liquid surfaces (Schmidt, 2007). Although increasing wind can increase surface roughness of a water body, the extent of convection also depends on wind direction and surrounding geography. Relationship of Measured to Predicted Fluxes In order to account for the background Se flux measured by the system, the average background flux (1.6 ng/m h) was subtracted from each measured flux value. This assumes the background flux rate is constant as opposed to the system measuring a constant background mass (and therefore not dependent on the time of sample). The implications of this assumption are insignificant, however, because subtracting the average background mass recovered gives an almost identical result to subtracting the background flux. In the majority of measurements, the background correction was small relative to the measured value. m2h 5124/m2h 10m mls significantly to­atmosphere m2h) 64 The majority of measured flux rates fell significantly below their corresponding predicted flux values (Figure 17). This was most clearly seen under the higher predicted flux condition driven by the higher volatile Se concentration of sample 3C (Table 14). Measured flux in this sample was an order of magnitude below the predicted flux (9.28 and 105.89 ng/respectively). The low measured flux relative to predicted flux indicated inefficiency in the flux measurement or inadequacy of the model to reflect volatilization in the Great Salt Lake. Inefficiency in Direct Flux Measurement To explore possible inefficiencies in the flux measurements, DMSe recovery was examined in tests ( described above) involving addition of small drops DMSe spike solution under the chamber. Results ranged from 7% to 24% of the input mass of Se (Table 15). The cause of low mass recovery is unknown, but a likely possibility is the adsorption of DMSe vapor to surfaces, which appears to be especially significant in absence of other vapors (e.g. water) that may compete with DMSe for surfaces. Inefficiency via loss of DMSe to surfaces is supported by calibration tests for the purge and cryo-trap system (described in Appendix B) which show a consistent 25% recovery that is attributed to partitioning of Se, primarily in the vapor phase, to various surfaces in the system. The loss of DMSe from the aqueous phase to surfaces appears to be low relative to loss from the vapor phase, as indicated by measurements of aqueous DMSe concentrations under controlled conditions (described below). The apparent lower recovery of the direct flux measurement system relative to the purge and cryo-trap system is consistent with the much larger surface area in former relative to the latter. Another possible contributor to the low recovery in the direct flux measurement system is m2h, ofDMSe 65 Predicted vs. Measured Se Flux 0 1 10 100 1000 Predicted Se Flux (ng/m2hr) Predicted vs. Measured Se Flux 0 50 100 150 200 2 Predicted Se Flux (ng/m hr) Figure 17. Relationship between measured and Estuarine model-predicted volatilization rates from the GSL. Measured volatilization rates include subtraction of average background flux of 1.6 ng/m h. Top: log axes. Bottom: linear axes. 1000~----------------------------------~ .-.1=0-. 100 Ne Oil = '-' ~ = ~ 10 c.l 00 "0 c.l 10. = '~" c.l ~ O~-------'---------r--------.-------~ o m 2hr) 200 180 'i:' 160 ..= N~140 = ';; 120 _ = ~ 100 c.l 00 "0 80 c.l 10. = '" 60 ~ c.l ~ 40 20 • 0 Predicted Se Flux (ng/m2hr) m2h. 66 Table 15. Results of attempted flux recovery test. Results show significant partitioning of volatilized Se to surfaces of the measurement system. Sample ID Mass Added US) Mass Recovered (MS) Percent Recovery SRI 0.0075 0.0018 24.3% SR2 0.0119 0.0008 7.0% SR3 0.0163 0.0022 13.5% (J.Lg) (J.Lg) 13 .5% the lack of water vapor due to addition of small drops (50 to 150 uX) to the system. The presence of water vapor in the purge and cryo-trap system may contribute to the higher recoveries observed in that system. Inefficiencies in the direct flux measurements were explored under controlled conditions that reflected the presence of water vapor in the system under field conditions. Fifty liters of GSL water was spiked with DMSe in a stainless steel container (described above) and the flux was measured over this container. Figure 18 (and Table 13) show the measured flux determined under controlled laboratory conditions (zero wind, constant temperature). Volatile Se concentrations ranged from zero in the background samples to 27 ng/L and were independently verified by the purge and cryo-trap system described in Appendix B. Measured fluxes exhibited a strong 1/10 linear direct relationship with the predicted fluxes under these controlled conditions, indicating that the flux measurement system was 10% efficient in measuring the actual DMSe flux. We conclude, that the 1/10 relationship between measured and predicted fluxes under controlled conditions is due to systemic inefficiency in the direct flux measurement, likely resulting from partitioning of volatile Se to surfaces in the vapor phase. Correcting Direct Flux Measurements for Background and Inefficiency To account for measurement inefficiency determined above, corrections were applied to the flux measurements taken on the GSL. In each sample, the measured flux rate after background subtraction was multiplied by 10 to correct for the 1/10 measurement inefficiency observed under controlled laboratory conditions. Figure 19 shows the corrected flux rates from the GSL. The majority of corrected measured fluxes are close to, but higher than the predicted fluxes. Two points (samples IB and 2C) are 67 J.lL) ngiL 1110 111 0 Inefficiency 1/1B Figure 18. Relationship between measured and Estuarine model-predicted volatilization rates under controlled laboratory conditions. Line represents the linear fit to controlled laboratory conditions. Controlled Environment Predicted vs. Measured Se Flux by Volatilization 50.---------------------------------------------------------~ 45 -;:- 40 r. N E 35 "- til .=.30 >< ~ 25 CII III 20 'a .C.II iil 15 III CII ::E 10 5 • Measured Flux = 0.1 x Predicted Flux R2 = 0.9893 O -~----------~--------~----------~----------~----------~ o 100 200 300 400 500 Predicted Se Flux (ng/m2hr) 68 69 Predicted vs. Measured Se Flux After Corrections Predicted Se Flux (ng/m hr) Predicted vs. Measured Se Flux After Corrections Predicted Se Flux (ng/m hr) Figure 19. Relationship between measured and Estuarine model-predicted volatilization rates from the GSL after inefficiency correction. Background flux was subtracted from measured flux prior to correction for measurement inefficiency. Top: log axes. Bottom: linear axes. 1000~----------------------------------~ 200 180 l:' 160 M..ec 140 Oil = ';' 120 = ~ 100 Q> 00 'tl 80 Q> ='"' '~" 60 Q> ~ 40 20 0 0 10 100 nglm250 100 150 nglm21000 200 70 significantly higher than their corresponding predicted fluxes. No obvious differences in conditions were observed during these two samples to cause this discrepancy. The high measured flux (after correction) relative to the predicted flux could result from a number of factors. One possibility is the underestimation of flux by the predictive model (Appendix C), which could potentially result from influences of the high salinity of the GSL. The air-water transfer velocity (kw) is inversely proportional to the Schmidt Number (Sc) to a power between 1/2 and 2/3. Sc is a dimensionless ratio of kinematic viscosity to molecular diffusivity, both of which are influenced by salinity. Unfortunately, the rate of change in each of these parameters as a function salinity at levels of the GSL could not be determined; hence, we can only suggest that the hyper salinity of the GSL may play a role in the discrepancy between measured and predicted fluxes. Another possibility is that the corrected measured fluxes are biased high relative to the actual flux rates from the GSL. At higher flux rates, accumulation of volatilized Se in the vapor phase in the chamber may occur, leading to a lower concentration gradient between the water and vapor phases, and resulting in inhibited flux. Since a majority of the laboratory flux tests using GSL water spiked with DMSe simulate higher flux conditions, the measured flux rate may have been inhibited to a greater extent than that which occurred in the lower flux conditions in the field. The result would be a slight over-correction of the field flux measurements from the 1/10 relationship between measured and observed fluxes in the laboratory. However, the laboratory tests under controlled conditions negate this possibility; since the measured flux rate was constant despite changes in sweep rate (see Methods and Table 13). ofthe ofthese playa 1110 71 A third possibility is that the differences between measured and predicted fluxes are magnified because of the relatively narrow range of fluxes that could be measured on the GSL. If multiple flux measurements could have been made under more turbulent conditions (i.e., higher flux), the discrepancy may have been reduced in significance. Though the exact cause of the discrepancy is unclear, the proximity of measured fluxes to predicted fluxes within the limited dataset of direct measurements leads us to conclude that a correction of the predicted fluxes is not warranted. The predictive model is an appropriate means of estimating annual removal of Se from the GSL by volatilization. Conclusion Accuracy in direct measurement of volatilized Se from the GSL was limited due to inefficiency in the sampling system. Calibration tests of the purge and cryo-trap system and multiple lab tests of the direct flux measurement system under highly controlled conditions lead us to conclude that 10% of actual flux is captured by direct measurement. Corrections were applied to measured fluxes for background flux and measurement inefficiency. After correction, most measured fluxes were close to, but slightly higher than predicted fluxes. Though the cause of this discrepancy is unclear, it is concluded that a correction of the predicted fluxes is not warranted. This investigation shows that significant volatilization of Se is occurring from the GSL and the predictive model applied by Diaz et al. (2008) is an appropriate method for estimating annual Se removal by volatilization. APPENDIX A CORE TRACE ELEMENT CONCENTRATIONS Core 2565-3 Interval cm 0-2 4-6 6-8 8-10 16-18 32-34 uq/q 56.24 59.59 58.39 51.65 55.65 46.96 42.41 34.28 37.90 29.74 Al uq/q 5390.43 5719.92 5394.15 4322.72 4746.10 4704.21 6341.17 4978.44 5969.63 3890.89 Ti uq/q 441.03 475.90 466.90 467.54 433.79 417.17 434.02 422.47 424.02 418.31 uq/q 16.14 17.19 16.66 15.39 15.04 13.75 15.37 13.84 15.43 12.55 Cr uq/q 8.08 9.16 7.58 6.04 7.09 6.93 8.64 7.94 8.93 6.25 uq/q 121.22 136.76 127.29 112.47 141.85 135.77 156.94 123.44 151.60 110.95 Fe uq/q 4428.18 4643.07 4090.65 3405.34 4116.57 3954.85 4735.59 3614.75 4351.25 3330.41 Co uq/q 2.40 2.26 1.91 1.65 1.94 1.85 2.01 1.63 1.92 1.47 Ni uq/q 7.22 6.93 5.45 4.31 5.85 5.50 5.80 5.17 6.48 4.30 UQ/q 65.75 14.46 7.16 5.49 6.46 6.03 6.71 68.19 7.00 93.51 Zn uq/q 64.63 52.87 25.70 21.73 23.95 24.92 26.28 21.35 25.31 20.76 As uq/q 13.82 10.10 9.69 10.12 10.87 11.11 12.66 9.93 10.86 9.89 Sr uq/q 2726.10 2422.37 2748.97 3022.20 2559.91 2522.21 2410.72 2522.28 2322.72 2926.35 Mo uq/q 16.22 2.90 2.17 1.47 2.76 1.34 1.63 1.58 2.24 1.75 Cd uq/q 1.06 1.17 0.30 0.17 0.20 0.22 0.20 0.24 0.19 0.23 Sb uq/q 0.54 0.42 0.15 0.14 0.18 0.14 0.22 0.15 0.15 0.13 Ba uq/q 318.55 324.49 380.75 352.00 362.26 292.86 315.11 300.22 354.61 491.22 TI uq/q 0.24 0.20 0.18 0.14 0.20 0.21 0.22 0.19 0.24 0.22 uq/q 82.85 84.58 9.41 5.76 6.01 6.75 6.84 5.90 5.74 6.50 U uq/q 7.87 4.56 4.38 4.45 4.26 3.87 4.08 3.93 3.87 4.45 Core 3510-BOX cm O-l 2-3 3-4 4-5 6-7 7-8 9-10 uq/q 151.98 114.83 106.04 77.32 66.96 39.36 49.52 57.84 68.58 67.12 Al uq/q 3329.93 9802.64 9807.86 9457.28 9878.15 5414.29 6272.58 7783.99 9125.40 9197.08 Ti uq/q 245.55 382.43 388.01 417.37 467.02 402.60 385.02 434.00 457.42 436.13 V uq/q 20.26 30.25 31.35 31.30 27.45 15.66 15.57 19.21 21.49 20.82 uq/q 11.56 17.00 20.30 19.40 17.51 9.91 9.84 11.62 11.91 12.38 Mn uq/q 239.27 301.06 305.70 295.16 231.80 136.42 147.41 181.48 212.07 216.90 Fe uq/q 6622.11 9707.80 10207.79 10431.44 9144.79 5420.27 5972.46 7204.78 8131.12 8125.69 uq/q 3.85 5.38 5.84 5.97 4.91 2.81 3.05 3.50 3.83 3.85 Ni uq/q 12.42 15.42 17.29 17.23 14.01 8.20 8.72 10.16 10.90 10.96 Cu uq/q 180.21 278.73 318.50 366.29 293.23 178.52 195.14 183.98 128.34 101.90 uq/q 107.05 146.69 163.26 164.60 129.35 81.48 93.96 110.28 97.16 85.89 As uq/q 32.93 38.02 33.50 36.57 33.59 19.41 17.67 17.87 15.07 14.19 Sr uq/q 359.25 700.28 844.22 1005.18 1261.32 964.15 1032.44 1258.98 1344.34 1370.39 Mo uq/q 44.18 57.01 66.87 78.21 47.00 30.28 16.33 16.01 11.93 9.42 Cd uq/q 1.38 2.55 3.75 4.92 4.52 2.42 2.12 2.03 1.47 1.26 Sb uq/q 1.51 0.99 1.31 2.50 1.71 1.29 1.00 1.67 1.70 1.19 Ba uq/q 146.35 226.74 256.29 283.53 266.80 159.15 175.87 227.73 274.50 288.75 TI uq/q 0.37 0.59 0.75 0.99 0.77 0.47 0.42 0.43 0.36 0.34 Pb uq/q 43.79 85.03 114.31 158.74 163.01 96.23 100.76 129.64 109.03 93.85 U m/q 3.15 5.44 6.53 7.15 6.04 3.38 3.42 4.19 4.25 4.20 Core 2565- 3 Interval em 0-2 2-4 4-6 6-8 8-10 10-12 12-14 14-16 16-18 32-34 Li lIa/a 56.24 59.59 58.39 51.65 55.65 AI lIa/a 5390.43 5719.92 5394.15 4322.72 Ti ua/a V ua/a Cr lIa/a Mn lIa/a 121.22 136.76 127.29 Fe lIa/a 4428.18 4643.07 Co lIa/a 2.40 2.26 1.91 Ni lIa/a 7.22 6.93 5.45 4.31 5.85 Cu lIa/a 65.75 14.46 7.16 5.49 6.46 Zn ua/As ua/a 13.82 10.10 9.69 10.12 Sr ua/a 2726.10 2422.37 2748.97 Mo ua/a Cd ua/a 1.06 1.17 0.30 0.17 0.20 Sb lIa/a Ba lIa/a 352.00 362.26 TI ua/a 0.24 0.20 0.18 0.14 0.20 Pb ua/a 82.85 84.58 9.41 5.76 6.01 6.75 ua/a 7.87 Core 3510- BOX Interval em 0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8-9 9-10 Li lIa/a AI lIa/a Ti lIa/a 245.55 382.43 388.01 417.37 467.02 402.60 lIa/a Cr lIa/a 11.56 17.00 20.30 19.40 17.51 9.91 Mn ua/a 239.27 301.06 305.70 295.16 231.80 Fe ua/a Co ~a/a 3.85 5.38 5.84 5.97 4.91 2.81 Ni ua/a 12.42 15.42 17.29 17.23 14.01 Cu ua/a Zn Ja/a 107.05 146.69 163.26 164.60 129.35 81.48 93.96 110.28 As ~a/a 32.93 38.02 33.50 36.57 33.59 Sr ~a/a Mo ua/a 44.18 57.01 Cd lIa/a 1.38 2.55 3.75 Sb lIa/a Ba ua/a TI ua/a 0.37 0.59 0.75 0.99 0.77 0.47 Pb ua/a Uq/q 3.15 6.53 73 cm uq/q 26.39 34.34 30.62 28.67 31.29 30.55 31.02 33.37 31.04 35.58 Al uq/q 6594.71 7609.45 6509.82 5319.88 7509.31 5729.27 5654.26 4589.44 5519.49 5955.36 uq/q 402.69 416.72 406.62 369.83 392.08 370.55 383.02 342.60 346.08 327.27 uq/q 16.05 15.95 14.19 12.98 15.15 12.83 13.30 11.20 12.14 12.72 uq/q 9.59 11.34 10.00 8.41 10.36 8.46 8.57 8.05 8.64 10.03 uq/q 325.88 289.51 245.75 254.76 293.63 279.26 265.47 275.76 299.49 328.56 uq/q 5616.95 5846.58 4952.47 4823.85 5577.63 4975.89 4689.60 4703.01 5174.14 5692.52 uq/q 3.62 3.04 2.65 2.46 2.78 2.51 2.55 2.72 2.76 2.95 uq/q 7.43 6.61 6.25 5.75 6.25 5.76 5.59 5.74 5.80 6.67 uq/q 31.23 6.45 4.69 4.80 4.88 4.84 4.38 4.69 4.74 5.79 uq/q 70.22 34.05 31.15 28.71 28.49 26.59 27.53 26.59 26.48 30.65 uq/q 23.42 20.84 23.85 23.27 25.77 24.06 24.33 26.09 24.94 21.47 uq/q 2475.51 2375.22 2698.93 2583.27 2262.63 2429.71 2642.27 2324.04 2132.94 1681.33 uq/q 19.20 9.96 7.09 6.62 9.56 8.27 9.15 7.47 5.59 4.43 uq/q 0.45 0.11 0.09 0.09 0.12 0.12 0.11 0.11 0.11 0.14 uq/q 2.19 0.42 0.33 0.25 0.41 0.33 0.40 0.39 0.45 0.38 uq/q 354.13 352.15 299.08 352.43 355.67 356.64 344.63 316.53 286.51 307.21 uq/q 0.24 0.16 0.17 0.12 0.16 0.18 0.17 0.17 0.14 0.16 uq/q 74.71 8.09 5.70 5.55 5.25 5.54 5.63 5.36 4.76 5.09 U uq/q 9.84 5.82 5.53 5.17 5.51 5.86 6.06 4.99 4.93 4.59 Core 2267-2 Interval em 0-2 2-4 4-6 6-8 10-13 13-16 16-19 19-22 25-28 84-88 Li ).lg/g AI lla/a Ti llOjO V ).lgjg Cr lla/a Mn lla/a Fe ua/a Co ).lojo Ni ).lg/g Cu ua/a Zn ).lo/o As Ug/g Sr ).lg/g Mo lla/a Cd llOjO Sb uojo Ba lla/a TI lla/a Pb ua/a llOjO APPENDIX B PURGE AND CRYO-TRAP SYSTEM Used with permission from Diaz et al. (2008) describing measurement of aqueous volatile Se concentration. Collection of volatile Se from the water involved a cryo-focusing trap system (Figure 20) following concepts used by researchers at the University of Pau in France (Amouroux and Donard, 1996). The system consisted of a reactor (a modified desiccator) with a diffuser connected to a helium line. The reactor sparges 7 liters of hypersaline water. The vapor swept from the reactor moved via Teflon tubing to a glass water trap (-55°C, dry ice/ethanol) to remove water from the flowing vapor. The vapor then entered a glass trap (-196°C, liquid nitrogen) to trap the volatile compounds collected from the water. Studies demonstrate that the entire volume of water can be sparged at a helium flow rate of 300 mL/min for approximately 15 min. After collection, nitric acid was added to the glass trap to oxidize volatile Se compounds and convert them to their stable aqueous species. The closed trap was digested in a water bath at 75°C for 3 hours, and the solution was analyzed for Se by ICP-MS at the University of Utah CWECS laboratory facility. The purge and cryo-focusing trap system was calibrated with dimethyl selenide (DMeSe) (AlfaAesar, 99% purity), which is reported to be the most stable volatile Se compound in seawater (Amouroux et al., 2000). APPENDIXB aqueous concentration. diffuser of300 mUmin aI., • dl Flowmeter & II « Helium tank Reactor Water trap (-55°C) ethanol/dryice Exit gas U glass trap C) liquid nitrogen system (Figure courtesy of X. Diaz, University of Utah) Flowmeter Watertlap ( -ethanoll dry ice ~."...--... Exit gas (-196°C) Ii qui d ni tro gen 75 Figure 20. Schematic representation of the volatile selenium cryo-focusing trap collection ofX. This system was tested in the laboratory using Great Salt Lake water spiked with pure dimethyl selenide. The analyzed spiked dimethyl selenide concentrations were equivalent to the expected value (within the 95% confidence limit) based on the calibration curve (Figure 21). Since measurements of pure water yield apparent volatile Se concentrations of 0.04 ± 0.01 ng/L, the practical detection limit for volatile Se using the purge and trap system is 0.04 ng/L. These results demonstrate that the system can quantify volatile Se concentrations in the nanogram per liter (ng/L or ppt) range. This resolution is 100 to 1000 times greater than typical analyses used for aquatic contaminants. It should be noted that the regressed recovery of volatile Se was 25% due to losses in the system. Therefore, measured values were corrected for 25% recovery according to the regression on Figure 21. The losses yielding the 25% recovery likely include partitioning to stainless steel, ceramic, glass and teflon surfaces in the chamber and tubing, and to epoxy sealant in holding the lid of the chamber (which was a modified dessiccator). Between samples, the entire system was thoroughly cleaned by rinsing five times with nitric acid (4 L, 2%) and deionized water (4 L). Tests demonstrated that volatile Se concentrations returned to background concentrations after cleaning. The calibration curve was used to correct the values measured in the field. Laboratory tests were run using pure water and Great Salt Lake shallow brine water with and without spiking of DMeSe to determine the analytical. This error was determined to be 13%, which includes the error associated with the ICP-MS analyses. 76 ± nglL, ngiL It and tubing, and to epoxy sealant in holding the lid of the chamber (which was a modified dessiccator). Between samples, the entire system was thoroughly cleaned by rinsing five times with nitric acid (4 L, 2%) and deionized water (4 L). Tests demonstrated that calibration curve was used to correct the values measured in the field. 77 Figure 21. Calibration curve for dimethyl selenide using the purge and trap system. (Figure courtesy of X. Diaz, University of Utah) 3.0 - 2.5 -..J Cl -t: 2.0 ~ ~ 0 > 1.5 Q) (J) "C Q) 1.0 ~ :s I/) co Q) ::!: 0.5 0.0 0 1 2 [Se]measured = 0.2416 [Se]expected + 0.04 R2 = 0.9662 -, • Spike in GSL water 3 4 5 6 Expected Se volatile (ng/L) 21 . ofX. 7 APPENDIX C FLUX CALCULATIONS Used with permission from Diaz et al. (2008) describing the volatile flux predictive model Models To estimate the volatile Se flux from the Great Salt Lake to the atmosphere, several models are available in the literature. These models have been used for estimating fluxes in fresh and seawater. The general equation for mass transfer flux for a volatile compound between two phases is defined in terms of the overall mass transfer velocity (kPhi/Ph2) and the concentration gradient between the phases (Schwarzenbach et al., 2003). An expression for the volatile Se flux in the Great Salt Lake is given below with the assumption that mass transfer is kinetically controlled in the water phase, as opposed to mass transfer in the vapor phase being the kinetically limiting process. Flux - akw C^f J r _ Cwtte°q)- aK r^VSe water 0VSe A K H G S L J (mol/mz/yr) (11) where: a is a unit correction factor (= 0.24); kw is the water transfer velocity in the air-water interface (cm/h); ^water is the concentration of volatile Se in water (mol/m ); ^water APPENDIXC at. model (kphllph2) aI., [ cVSe J _ VSe VSe,eq _ VSe air Flux - akw (Cwater - Cwater )- akw Cwater - K 2 HGSL m /(11 ) kw air-c~ 3 c~~ water water m); water K"HGSL = 0.0248 exp(0.0418T)^ 0**[sa/flto( (12) where Ks is the salinity constant, and [salt]t ot is the total molar concentration of salt. Dimethyl selenide (DMeSe) is the most important volatile Se compound found in air; and in fresh and saline waters (Atkinson et al., 1990; Neumann, 2003; Tessier et al., 2003), and therefore is an appropriate species on which to base our estimations. The Ks for 3 pVSe is the equilibrium concentration of volatile Se in water (mol/m ); ^ a i r is the K concentration of volatile Se in air (mol/m ); HGSL is the dimensionless Henry's constant for volatile Se for the Great Salt Lake. In our case, concentrations of volatile Se in the water have been measured. Concentrations of volatile Se in the air can potentially be measured; however, in the estimations below we assume this concentration to be zero. Dimensionless Henry's constant correction The dimensionless Henry's constant ( HGSL ) and the water mass transfer velocity in the air-water interface (kw) were determined using empirical models from the literature. These models are based on wind velocity, water temperature, viscosity and diffusivity of the volatile species. The viscosity, diffusivity, and dimensionless Henry's constant each require corrections for the salinity of the Great Salt Lake, which is 3-5 times greater than that of the ocean. An equation to estimate the dimensionless Henry's constant for DMeSe as a function of temperature was developed by Guo et al. (2000), whereas a salinity correction was provided by Schwarzenbach et al. (2003), yielding: 79 is the equilibrium concentration of volatile Se in water (mol/m\ C~~e is the concentration of volatile Se in air (mollm\ K~Ic;SI is the dimensionless Henry's constant for volatile Se for the Great Salt Lake. In our case, concentrations of volatile Se in the water have been measured. Concentrations of volatile Se in the air can potentially be measured; however, in the estimations below we assume this concentration to be zero. (K~GSI kw) KH" = 0.0248 exp(0.0418T) * 1 OKS [saltl tot GSL (12) KS tot KS 80 kw(cm/h) = f V1/2 Sc 600, (2 + 0.24*4) foruio>5m/s (13) kw(cm/h)= (2 + 0.2414) V D U U y foruio <5m/s (14) where Sc is the Schmidt number, and " I O is the wind velocity measured 10 m over the surface of the lake. Saltzman et al. (1993) defined a Schmidt number for DMeS as a function of water temperature (°C) and corrected for the sea water salinity (via coefficients) as follows: ScSDeMeTer = 2674.0-147.127 3.726V2-0.0387"3 (15) Water transfer velocity - modified Liss and Merlivat model An alternative approach is provided by the modified Liss and Merlivat model (Livingstone and Imboden, 1993; Liss and Merlivat, 1986), the results of which largely DMeSe was not available from the literature, whereas a value for dimethyl sulfide (DMeS) was available, and was used on the basis of its similarity to DMeSe (Amouroux, 1995). Water transfer velocity - Estuarine model To calculate the water transfer velocity, an approximation used in the Hudson estuary by Clark et al. (1995), corrected for the Schmidt number according to the boundary layer model (Schwarzenbach et al., 2003). This so-called Estuarine model is as follows: aI. aI., ( J -1/kw(cmlh)= :a~ (2+0.24u;o) for UIO > 5 mls