7—... v —_‘.. .nvxw , r#;'wmww=r A _ ."‘ J.) ., ~M- .‘ _. . p, -A- - -., . 1;; M‘ .... -. 2‘... ' 2 $ . 353' " 1|! \ 2,23 ‘... ..2 . In“ ' ~o‘ -u.-pC:~‘ o ..-‘ Jmu-- um;- 4 2- a -.‘ - , 2 ‘4’; O o.‘ 'Wu . -— ”W q ....—".’.‘.‘3‘: ...._.. Mm - . ‘1 "San: l.- - . 1:32... .J‘ ‘.. ’ 'v , h..— v- “P: vuw .. w.~—~.._. .Mza . -4: ~ . ‘q “ “I . I“ ‘ u. s .- ‘q-r _ .4 ~— h. a "—3 -— a -0 .v' '\ .3573..2'2t ' 43132 3 .- "q ‘f 11' . J 2143433922231 3332424332 ' I“? 72 I .51.‘-H35;',7.‘32-:-;:§‘!: ‘ $33334“; "22;? I; 5‘ :13‘1-‘.[.11}2.2;,-,13{233‘ ' {lei-1!! :15“ 31' 2.‘ 22- 3 21k! 5.733321% 23’32’ ,. .22.. 32% 32. 24‘2222. 2.22.23.2-2322 2 3' J3 "'2'2'222222522». a,” .2',fl .2146qu ,|' 3‘2 ”92' $23,325;- 4333' 27",} 34233483! T 33‘3330‘ 313113? :3. 2233 In! 3‘ 421‘} ’ '2} 33"42’32 '4’}"""2233 332:2 “33,413 . 2232’ '4 '2} ‘ {£223} ;-.‘ LL“! hi fi- ’ 32457322353? 2'. '3‘: 345‘- 4 £3 3 '34“. .3331 ara‘f‘l‘i ”$533. z -;2 .12 -..2 23-! 2. ‘ .22243234'3333 4443-4 42.333 2. ‘4‘. 33151;” '3 344 3i ~ ;:‘:‘i. :33”: dzfiffi 2‘ ‘ " ,- v-vh...‘ ' ' ~J"“ v {fiv-va- . o .o.,. n. 0.. " . . u- a 4 ‘ ‘ ' . a . . . ' .- . 23'; 'V 5331'! :3‘: . ~3,r:: A4 2 "3459-: t‘ ' ..-2, ,r?‘.$?:33z ,6.’ 'T-l.""zl4" :A::p v H 51“”3225 . 3 “ “4-42" 4 "2;,2'2'2‘ < ‘ '1 3 :" '-. ”‘72.! . 4 v ' .3 .I. 2.2.232 273.222, .'. 3 :34 4~ 32-? - I?" 331434!“ ‘2}!!!3'33 Sing a. 3‘ v ., 333‘; “1'13 {"0 gut”: €4,323; 3' .fi'l {111.5, 3:“ ‘L ' 3‘, ”I “’11“ of fi‘J’ X‘fi J12 (’4’? 4.14 2.2.1,!” 35ft1321£§3§fp+$§bfi1 3'3 W313", ‘3 «2132; "‘ 33 3 “"3924 £133 35433:. 1"3"“‘933352333’4‘433‘11'43’ Wyn“ 15%;. 3334!“: I333. 3‘ “3 {£1 3‘33 22323214315 4! if}, 3133 ”(W'Lm‘ at; 4 2,41 2222 .2222! 3,1332 {322.1232} fie "£37., 2., 2 '2 u If 2‘11” \ ._.“ ‘0!" n ‘ ~ I 38‘4””?322 (3‘32 ' L. 33;? £3433 3' 3 [m 1.1;?“ I! 7331'! 3.3%.va fixfll‘ Q, 3::g' ‘33}? 12. 2,} 1*, 322222.22 (23 323 3232.%“2~2w232{2 ' 3 2. «33" “$2331. 4 2:343fi52132i41fi3‘! g\ 3 25222222232223 11232 22222233223232 22232223313122.“ 2332:” 22222.15 ‘2 u 1. I $0540 9‘ 8"?— ( ST TE UNIVERSITY LIBRARIES“ llll llllllllllll Illlll llll l l 3 1293 00533 0356 LISRARY Michigan State University This is to certify that the thesis entitled The Chemical Behavior of Heavy Metals at the Water-Sediment Interface of Selected Streams in Maine Based on Ternary Partitioning Diagrams presented by Dale Henry Rezabek has been accepted towards fulfillment of the requirements for Master of Science degree in Geology Major professor Dr. David T. Long Datefié “IOU 8? 0-7639 MS U is an Affirmative Action/Equal Opportunity Institution MSU RETURNING MATERIALS: Place in book drop to LJBRARJES remove this checkout from .—:_-—. your record. FINES will be charged if book is returned after the date stamped below. {SQ C" 513%? MAGIC 2 i wit/$1998 THE CHEMICAL BEHAVIOR OF HEAVY METALS AT THE HATER-SEDIMENT INTERFACE OF SELECTED STREAMS IN MAINE BASED ON TERNARY PARTITIONING DIAGRAMS BY Dale Henry Rezabek A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Geological Sciences 1988 ac- de* bel in tre cor The ext cor dla beh dis tho of . ads. abm tecl 610: 5/3 0/0577 ABSTRACT THE CHEMICAL BEHAVIOR OF HEAVY METALS AT THE WATER-SEDIMENT INTERFACE OF SELECTED STREAMS IN MAINE BASED ON TERNARY PARTITIONING DIAGRAMS BY Dale H. Rezabek The many chemical and physical controls on heavy metal activities in natural aqueous systems have impeded the development of a quantitative model for predicting metal behavior. This research studies the partitioning of metals in stream sediment adsorbents to determine if systematic trends in metal adsorption behavior are reproducible and consistent with adsorption theory and experimental results. The techniques need are: sequential selective chemical extractions, normalization to equal adsorbent concentrations, and the plotting of data on ternary diagrams. The results show: 1) unique fields of metal behavior can be depicted on the ternary diagrams, 2) the distributions of naturally-added metals are consistent with those of anthropogenically-added metals and with the results of adsorption studies, 3) the absolute abundances of adsorbing phases are apparently not as important as relative abundances in controlling metal partitioning, and 4) these techniques do not quantity the data sufficiently to be used alone in predicting metal behaviors. ACKNOWLEDGMENTS I would like to express my sincere appreciation to Dr. David T. Long, my graduate advisor and committee chairman. His help was invaluable in completing this research project. I would also like to thank my graduate committee members, Dr. John T. Wilband and Dr. Michael A. Velbel, for their assistance in my research. My appreciation is extended to my fellow geochemist, Timothy Wilson, who became a good friend and colleague during my graduate studies. I thank Michael Miller for his friendship and advice. Finally, I would like to especially thank my best friend and wife, Margit, for her love, understanding, and undying faith in me. 11 Chapt Ce Chapter Chapter Fi La Co Chapter D11 le le Met Chaptfir TABLE OF CONTENTS EASE Chapter One: Introduction ................................ 1 The Problem.... .......... . .......................... 1 Metal Partitioning in Sediment ...................... 3 Phase Concentration Factor .......................... 5 Past Work by Gephart (1982) ......................... 8 Hypothesis ......................................... 13 Chapter Two: Description of the Study Area .............. 15 Jackman Township Area .............................. 17 Calais Area... ..................................... 20 Topsfield Area ..................................... 20 Chapter Three: Selective Chemical Extraction Theory and Application ........................... 24 Chapter Four: Methods ................................... 37 Field Methods ...................................... 37 Laboratory Methods ............ . .................... 40 Construction of Ternary Diagrams ................... 43 Chapter Five: Results and Discussion .................... 48 Differences in streams sampled in Maine ............ 5? Differences in areas sampled in Maine... ........... 61 Differences in substrate concentrations ............ 70 Metal behaviors and past research .................. 86 Chapter Six: Conclusions ....................... . ........ 90 iii Appe #e A: VIU Appendix 1: Geochemistry of Heavy Metals ................ 93 Appendix 2: Adsorption Theory and Experimentation ....... 98 Appendix 3: Tabulated Metal Concentration Data and Stream Water Data.... .................. 113 Appendix 4: Statistical Data for Phase Concentration Distributions for the Grand River, Michigan and for Streams in Three Areas in Maine......121 List of References ................. . ................... 123 iv F1 5‘1 Fl FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE LIST OF FIGURES Ease l: A summary of the results of Gephart (1982) ........................... 11 2: Study area locations in Maine ............ 16 3: Jackman Area Map ......................... 19 4: Calais Area Map .......................... 21 5: Topsfield Area Map ....................... 23 6: Flow chart of chemical extractions for this study ............................... 35 7: Example calculation of PCF value for sample Tom-6..... ............... . ........ 45 8: Ternary diagram for chromium partitioning in Maine sediments .......... 49 9: Ternary diagram for copper partitioning in Maine sediments .................. .....51 10: Ternary diagram for nickel partitioning in Maine sediments....... ...... . ...... ..53 ll: Ternary diagram for lead partitioning in Maine sediments..... ..... ............54 12: Ternary diagram for zinc partitioning in Maine sediments ...................... 56 13: Comparison of metal partitioning in fast-flowing and slow-flowing streams...59 l4: Chromium ternary diagrams for three areas in Maine .................... 62 15: Zinc ternary diagrams for three areas in Maine..... ........................... 64 16: Nickel ternary diagrams for three areas in Maine .................... 66 FIGURE 17: Lead ternary diagrams for three areas in Maine.... ....... ........ ...... .. ..... 67 FIGURE 18: Copper ternary diagrams for three areas in Maine... ................. 68 vi Ti TA TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE TABLE 10: 11: 13: ' Tessier et a1. LIST OF TABLES BBQ: . Terminology for sediment fractions ......... 4 Chemical phases and extraction methods....27 (1979) extraction methods..34 Summary statistics for raw phase concentrations and log phase concentrations in Grand River sediments ........... .......71 Summary statistics for raw phase concentrations and log phase concentrations in Maine sediments ........................ 73 Summary statistics for the raw phase concentrations for each of the three areas sampled in Maine.. ...... ............ ANOVA for Maine log oxidizable phase concentrations............... ANOVA for Maine reducible phase log moderately concentrations ............ 75 ANOVA for Maine reducible phase log easily concentrations ............ 75 Kruskal-wallis concentrations analyses for log phase in Maine sediments ........ 77 Comparison of sediment phase concentrations for Maine and the Grand River, Michigan ................ 79 Comparison of sediment phase concentrations for three sampling areas in Maine... ..... ...... ...... . ............ 81 List of samples with high concentrations of copper (> 100 ppm) in the hydromorphic phases...... ..... . ......... .83 vii in Th fo 5‘1 co pr ad Th pr su su an ex ad Co be an 19 he on as di CHARTER_QNE. lntrndnstinn. This study examines whether there are systematic trends in metal adsorption on sediments in natural aqueous systems. The purpose of this type of research is to develop a model for making predictions of metal behaviors in exogenic systems. Wm At the water-sediment interface, trace metals are controlled by kinetic and thermodynamic components of precipitation-dissolution, oxidation-reduction, and adsorption-desorption reactions (Stumm and Morgan, 1981). The chemical speciations of dissolved metals to predict precipitation-dissolution reactions have been interpreted successfully with the development of thermodynamic models such as VATEQZ (Ball et al., 1978), SQ3NR (Wolery, 1983), and MIHEOL (westall et al., 1976). Laboratory experimentation has led to the development of models for the adsorption of metals on single substrates under well-defined conditions, which can theoretically predict the adsorption behaviors of metals in simple natural systems (Balistrieri and Murray, 1983; Davis and Leckie, 1978; Gadde and Latinen, 1974; Leckie et al., 1980; Lion et al., 1982). However, natural systems are usually quite complex and have more than one adsorbing substrate. Therefore, integrated models such as HINTEQ (Felmy et al., 1984), which combine precipitation- dissolution and adsorption-desorption, are tenuous. At 1 tl di 5? co th Va; Cor 3Y5 equ 3Y3 9t ads the 0f 1 Varj tern Cale 2 present, the fate of a metal at the water-sediment interface can only be described qualitatively. One way to overcome the present lack of a useful model would be to derive the equilibrium constant for the adsorption of each metal on an adsorbing substrate or phase (Oakley et al., 1981). These constants could then be used to study metal adsorption distributions similar to the way thermodynamic equilibrium constants for precipitation- dissolution reactions are used for calculating metal speciation in the chemical modeling programs. However, it may be difficult to use this technique since adsorption is controlled by the system pH, pe, ionic strength, etc., and the constant would need to be adjusted (if possible) for all variables. Tessier et al. (1985) calculated apparent equilibrium constants for the adsorption of Cd, Cu, Ni, Pb, and Zn onto a natural substrate, iron oxyhydroxide, collected from lake systems in Canada. These results compared favorably with equilibrium constants obtained from simple experimental systems reported by Balistrieri and Murray (1983) and Oakley et al. (1981), and with theories on iron hydroxide adsorption (Leckie et al., 1980). The differences between the calculated constants for certain metals and the results of the experimental systems were interpreted to be caused by variations in the pH's of the systems and the formation of ternary complexes which were not accounted for in the calculations. Clearly, the complexity of natural systems me set tht Th: and org Hit cop “it. Get; Strl 1975 thro 3 necessitates a more comprehensive study of metal behaviors in multi-substrate systems before calculations like those made by Tessier et al. (1985) can be reliably made. This study explores a different approach to the development of a quantitative model for predicting metal adsorption in natural systems, by looking at natural metal behaviors on ternary diagrams after normalizing adsorption data to equal amounts of substrate. W Metals added to the sediment from the water (or those metals that can be easily released to solution from the sediment by environmental alterations) are associated with the hydromorphic fraction of the sediment (Gibbs, 1977). This fraction is composed of clay minerals, carbonates, iron and manganese (and other metal) hydroxides, sulfides, and organic matter (Gibbs, 1977). The association of the metals with these chemical phases (or substrates) is by adsorption, coprecipitation, complexing, or ion exchange of the metal with the phase. Non-mobile metals (frequently called the detrital or residual fraction) are found within the lattice structures of clay and silicate minerals (Gupta and Chen, 1975). Table 1 introduces the terminology which will be used throughout this report. Adsorbed-metal concentrations can be measured to help understand the factors that control metal exchange on hydromorphic phases of the sediment. One method of measuring metal concentrations associated with hydromorphic phases is 4 Table 1: Terminology for sediment fraction. Sediment Sediment what's Hydromorphic Detrital Clay minerals Carbonates, Cr-ox, etc. Mn-ox, amorphous Fe-ox Fe-ox Organics, sulfides Silicate minerals Chemical Phase. Exchangeable Acid soluble Easily Reducible Moderately Reducible oxidizable Residual to al se 19 se of se: Te: att met to bon- qua con Alt du: sel 3Urf °f t. 1UVO: 5 to define the sediment substrates in terms of chemically alterable phases. This can be determined in a series of sequential selective, chemical extractions (Luoma and Davis, 1983). A number of studies have utilized sequential selective chemical extractions to determine the partitioning of heavy metals associated with the hydromorphic phases of sediment (Chao, 1972; Halo, 1977; Gatehouse et al., 1977; Tessier et al., 1979). In theory, a specific reagent will attack one chemically-reactive phase and release any bound (adsorbed, coprecipitated, complexed or ion exchangeable) metals into solution. If reagents are applied in a sequence to release metals from the weakest-bonded to strongest- bonded of the different phases, they may provide a quantitative method for determining adsorbed metal concentrations from these "operationally defined" phases. Although there can be problems with readsorption of metals during the attacks (Rendell et al., 1980) and limited selectivity of chemical for a phase (Forstner and Patchineelam, 1980), the method of sequential selective, chemical extractions remains an important tool in the development of quantitative models for metal behaviors in sediment-water systems (Luoma and Bryan, 1981). W: In natural systems, substrates have wide ranges of surface areas. One method of study to overcome this problem of the variable surface areas available for adsorption involves the proportionating of trace metal concentrations in bi. ph: 'l'h COl 355 Va) wit 8t ”he; Con< Of a high meta 6 according to the concentrations of the defined phases (Filipek and Owen, 1979). Calculating the ratio of a metal concentration to a substrate concentration provides information on the competition of substrates to adsorb metals, on competition among different trace metals for binding sites, and for characterizing the chemical and physical processes that control metal partitioning behavior. This would be very difficult using only the total concentrations of adsorbed metals and adsorbing substrates (Nowlan, 1976). An estimate of the relative importance of the association for a phase with a metal can be made in two ways: 1. by calculating the ratio of a metal concentration with the concentration of the adsorbing substrate (Filipek et al., 1981), or 2. by calculating a phase concentration factor (PCF value) which is a ratio of the percentage of a metal among the phases to the percentage content of a respective phase within a sample (Forstner and Patchineelam, 1980). This can be shown in equation 1: PCF . M/MT divided by P/PT (l) where M s concentration of a metal in a phase, MT = total concentration of the metal in all phases, P = concentration of a phase, and PT = total concentration of all phases. A high value for the ratio indicates a high association of the metal with the phase. 7 Carpenter et al. (1978) looked at relative associations of metals with iron oxides and manganese oxides by using a normalization ratio shown in equation 2: (""n-ox / MFC'OX) diVided by (”nNn_ox / FeFe_ox) (2) where "Mn-ox = metal concentration in Mn-oxide extraction, "Fe-ox 8 metal concentration in Fe-oxide extraction, "nMn-ox 2 Mn concentration in Mn-oxide extraction, and FeFe—ox = Fe concentration in Fe-oxide extraction. Chemical extractions of sediment samples were used to distinguish between metals adsorbed on Mn and Fe oxides. By using this ratio technique, Carpenter et al. (1978) found Fe-oxides had a high association with Zn and Ni, and Mn- oxides had a high association with Cu, upstream from a mineralized zone. Downstream from this zone, high associations with Zn and Cu by these oxides were reversed. Pb was associated mostly with Fe-oxides both upstream and downstream from the mineralized zone. To detect anomalies in metal concentrations among iron and manganese oxides, a number of authors have suggested the use of ratios of metal concentrations to Fe and Mn concentrations (Tessier et al., 1982; Nowlan, 1976; Robinson, 1982). Filipek and Owen (1979) normalized metal concentrations to the weight of sediment dissolved in a given extraction rather than the total sediment weight. This was done to differentiate metal inputs from detrital St it 5E in Cd 9)! 8 components, which would represent nonmobile forms of metals in the sediment unavailable to the environment. Luoma and Bryan (1981) used a statistical filtering method for correlation coefficients to investigate competition of substrates for metals and to see if substrate concentrations in the sediment influenced that competition. The filter functioned by isolating substrates with strong correlations with metals. If two substrates (Fe-oxides and Hn-oxides) competed for a metal, the substrate with a low concentration would show higher correlation with the metal. High concentrations of substrates would be filtered out to see, for example, if correlations between Fe and metals improved (meaning that Fe had a higher association with the metal). This study was interesting, but did not utilize sequential extractions, which have been shown to be very important in characterizing metal partitioning by Tessier et al. (1979). Therefore, the results of Luoma and Bryan (1981) cannot be compared to other investigations using chemical extractions. WW Tessier et al. (1979) developed a technique for releasing metals from the following chemical phases (and speculated on the chemical nature of the phases): exchangeable (weakly adsorbed metals), acid soluble (carbonates), reducible (Fe-Mn-oxides), oxidizable (organic matter and sulfides), and residual (silicates). These C? Te Ca Ge Cu La in 0p» By: 361 the Cor (ex 0rd 333' am mat 613. the eas‘ OXId the 9 chemical extractions will be discussed in detail in Chapter Three. Gephart (1982) modified the procedure developed by Tessier et al. (1979) by splitting the reducible phase into easily and moderately reducible phases. With this procedure Gephart (1982) studied the partitioning behavior of Cr, Pb, Cu, Ni, and Zn adsorbed on the sediments of the Grand River, Lansing, Michigan. The Lansing area is highly populated and industrial. Automobile assembly, metal plating, and tanning operations are major sources of metals to the riverine system. The study concentrated on metal partitioning in sediment substrates; Mn-oxides, Fe-oxides, and organic matter. For most natural, oxic systems these appear to be the most important substrates (Gibbs, 1977). The concentrations of metals associated with other phases (exchangeable, acid soluble, and residual) were measured in order to distinguish them from the concentrations of metals associated with the three most important phases. Since it is difficult to quantify the concentration of a metal associated with Mn-oxides, Fe-oxides, or organic matter, a more general identification was used (and will be also used in this study) based on the chemical response of the substrates to the chemical extractions; respectively easily reducible (ER), moderately reducible (MR), and oxidizable (OX) phases. Gephart (1982) used a modification of the PCF to study the partitioning of the metals among the three substrates. 10 The modification was in the calculation of substrate concentrations, which are estimated from the chemical extractions. For example, ER phase concentration was estimated from the concentrations of Fe (dissolved amorphous Fe-oxides) and Mn (dissolved Mn-oxides) in the extraction solution after the chemical attack. In the report by Gephart (1982), this calculation for ER phases was done incorrectly, with the wrong Mn concentrations used. The corrected PCF values are calculated as percentages and plotted on ternary diagrams (Figure 1). These plots represent the partitioning of the metals relative to the substrates being present in equal concentrations (i.e. these are normalized metal data). In multi-substrate systems, if the relative masses of the adsorbing substrates alone controlled the partitioning behaviors of metals, then the normalization of partitioning data for a metal to equal amounts of adsorbing substrates present in the systems would be expected to plot as a point on a ternary diagram. Predicting metal partitioning in a natural water-sediment system could then be done by estimating substrate abundances from the results of sequential selective chemical extractions. However, the data for each metal in the study by Gephart (1982) are scattered, and this suggests that the assumption of a one to one change in adsorption as a function of change in the mass of adsorbent is not valid and may not by itself explain why the data is scattered. Other variables such as surface area of adsorbent, solution pH, pe, ionic strength, etc. may cause -:‘~ ER ER 11 0X I ox “ CR MR cu MR ER OX MR Nl ER ' ER ox MR ox MR ZN PB Figure 1: A summary of the results of Gephart (1982). 12 the scatter and need to be considered in modeling the behavior of metals at the water-sediment interface. In addition, this technique of normalizing partitioning data may be in error and needs study. The normalized partitioning data plot in unique clusters for each metal. For example, in Figure 1, note that the normalized data for Ni, Pb, Zn, and Cu plot in clusters trending away from the ER phase apex. These clusters appear to rotate around the ER apex away from the ER-OX phase boundary in the order Cu < Pb < Ni < Zn, with the Zn data clustering along the ER-MR phase boundary. Four general metal behaviors can be interpreted from the diagrams: 1. Except for Cr each metal plots in a fairly tight cluster, 2. Cu plots in a cluster essentially along the ox-ER phase boundary, 3. Cr plots in a scatter with most of the points clustered near the center of the diagram and some plotting near the ER-MR phase boundary, and 4. Ni, Pb, and Zn cluster in the upper part of the diagram along the ER-MR phase boundary. The data handling techniques developed by Gephart (1982) (Fig. 1) appear to depict unique metal behaviors, but it is not known if this type of diagram depicts similar behaviors of the metals in all water-sediment environments. That is, if adsorption data are handled in the same way experimentally and numerically, will partitioning patterns of metals on ternary plots such as Figure 1 be the same for various environments? These patterns may be artifacts of b. 1: me or te an 3? th th Pa: Sh< ads 13 anthropogenic metal sources. Therefore, this study will examine the application of these techniques in an area with natural additions of metals to streams. W This research examines whether similar trends as found by Gephart (1982) exist for metal partitioning in the different sediment adsorbents in natural aqueous systems. If this is found to be the case, then metal adsorption behaviors among substrates can be studied by normalization techniques in which the relative partitioning of a metal is identified rather than the absolute partitioning. This would be significant because this might provide a technique for predicting quantitatively the adsorption behaviors of metals in sediments. The hypothesis for this study is that the normalized metal distributions among the phases ER, MR, and OX depicted on ternary diagrams that were determined by the above techniques reflect systematic trends in metal behaviors among adsorbing substrates for any water-sediment system. By systematic, it is meant that there is some universality in the way these metals are associated with the adsorbents. If this hypothesis is true, then these metal distribution patterns should be found in different aqueous systems and it should also be possible to interpret the data in terms of adsorption theory and the results of laboratory experimentation. The test was performed in this study by applying these techniques of chemical extractions, . 14 normalization to equal amounts of substrate, and ternary diagram construction to sediments from streams located in areas where natural sources of metals to streams are greater than anthropogenic inputs. The process of normalization was also evaluated to determine whether its use in studying adsorption behaviors is actually advantageous for developing models. hr; wit ant dis nat mea sea sit 196 mm 21:) p10 9&0. Now map: Shm Area Calé deSc COnt kilo ”Ort; W W The criteria used for choosing an area to test the hypothesis were that the area should be a stream environment with active sedimentation, that the area should be free from anthropogenic sources of metals such as industrial waste discharges or landfills, and that the area should have a natural source of metals in adequate concentrations for measurement with an atomic absorption spectrophotometer. A search was conducted in the literature to locate potential sites that were accessible and met the above criteria. Two USGS Mineral Resource Maps of Maine (Post and Hite, 1964; Post et al., 1967) were located which identified a number of remote areas with high concentrations of copper, zinc, and lead in stream sediments. Concentrations were plotted throughout the state based on a multi-year geochemical survey of over 2000 Maine streams. A study by Nowlan (1976) shows that three areas in particular on these maps meet the above criteria quite well. These areas are shown in Figure 2 and will be identified as Jackman Township Area, Topsfield Area (which includes Tomah Mountain), and Calais Area. Some background data on the state of Maine is described below. Maine is the most northeastern and eastern state of the continental United States, approximately 80,277 square kilometers in area. Based on the average climatic data for northern and southern Maine from 1939 to 1979, Maine has a 15 hJEVV BRUNSWlCK TOPSFIELD JACKMAN CALAIS liTL.AhFTlC OCEEAP4 Figure 2: Study area locations in Maine. year mini zesp cm, 1981 so t to a of l hill The Spit inc] fore gene Prec 10ck 17 yearly mean temperature of 5.6°C (42°F), with a yearly minimum and maximum of o.s°c (32.9°F) and 10.6°C (51.1°F), respectively. Annual precipitation has an average of 99.8 cm, with an average snowfall of 238.5 cm (Ruffner and Bair, 1981). Southeastern Maine is bordered by the Atlantic Ocean, so the topographic relief ranges from 0 meters (sea level) to approximately 1525 meters above sea level (ASL). Classes of land-surface forms include plains with hills and high hills, open high hills, and open low mountains (USGS, 1970). The major vegetation of Maine consists of Northern hardwood- spruce forests (Genera include Agar, fistula, Eagus, Eigea, ragga), with some areas of Northern hardwoods (Genera include Anal; fistula. Enema. Tansa) and Northern spruce-fir forests (Genera include Eigga, Abiga). The geology (in general) of Maine comprises eugeosynclinal deposits of late Precambrian to mid-Paleozoic age, with intrusive granitic rock of late Paleozoic age in some areas. Although bedrock is exposed in many areas, there are many surface features throughout the state which show evidence of Wisconsinan glaciation, such as eskers, moraines, beach, glaciofluvial and glaciomarine deposits, and various deposits of till (USGS, 1970; Thompson and Borns, 1985). Jacknan_122nahin_Area The surficial geology of this area (570 sq. km.) consists of hilly terrain with elevations ranging from 270 to 1130 meters ASL. Moderate to thin glacial drift (< 3 18 meters thick) covers the area, with many bedrock outcrops. Most exposed bedrock has a thin cover of soil and vegetation. It has many low, swampy areas with cedar bogs. Hills are covered with spruce-fir forests. The only anthropogenic activity in this area has been some logging operations. The bedrock is Ordovician quartz monzonite of the Attean Formation and fine- to medium-grained gabbro, and Devonian slate, metasiltstone, and metasandstone of the Seboomook and Tarratine Formations. Some of the mineralization noted by Nowlan et al. (1983) include pyrite (Fesz), chalcopyrite (CuFeSz), galena (PbS), sphalerite (ZnS), and cobaltiferous-nickeliferous pyrrhotite (Fe(Ni,Co)S). Eight streams and tributaries were chosen for sampling in the Jackman Township area. Although some have names recorded on maps, others were unofficially named by the USGS during geochemical surveys. The streams, with widths ranging from 1 to 10 meters, are Pyrite Creek, West Pyrite Creek, Bean Brook, Parlin Brook, Chase Stream, Dead Stream, Cold Stream, and Alder Stream (Figure 3). Some streams that were sampled were in boggy areas, had high amounts of organic matter, and shallow depths as they flowed over glacial deposits. Other streams had rocky beds with boulders and exposed bedrock, and a few waterfalls. The streams in boggy areas have extremely high concentrations of manganese oxides in the form of black coatings on cobbles and boulders, and as discrete nodules in some places. This is prevalent in JACKMAN TOWNSHIP. WESTFORKS AREA fl ©® ' Prune CREEK Ll} “ 37H“ MISERY\ ’ is cl STREAM fi‘ ‘1. 3' / O 5' 6 '0 ' 3' e 3 ) H ‘\B£AN BROOK : ‘1‘ 1%}3 PARuu fl pano PARLIN\ ' . 13 ,, 1‘32 anoox ' . ‘3 COLD I? CHASE ” . STREAM , .. STREAM 5 \. tho 1 ,1 i . ‘8 N ‘ OEAo , l? STREAM M ‘/,K£NNEBEC RIVER WEST FORKS l. 0 2 mllu = , IV Figure 3: Jackman Area Map. 20 Maine for streams that drain swampy areas (Canny and Post, 1964). minim This area (124 sq km) consists of relatively flat terrain with a few hills and elevations ranging from 45 to 150 meters ASL. There are numerous swamps, bogs, and lakes in low areas, surrounded by Northern hardwood-spruce forests (USGS, 1970). The surface geology of thin drift deposits also has many outcroppings of bedrock, and some fine-grained sediments of glaciomarine origin (Thompson and Borns, 1985). The bedrock in the areas sampled consists of Devonian intrusives (granite, quartz diorite, and ultramafic rock). Four streams and tributaries were sampled in the Calais area: Eastern Stream, Western Stream, Mill Brook, and various branches and tributaries of the Magurrewock River in the Moosehorn National Wildlife Refuge (Figure 4). The streams had widths ranging from 1 to 5 meters. All the streams, except for Eastern Stream, had rocky beds with numerous bedrock boulders and cobbles. Eastern Stream (at the site sampled) flowed over a large, tabular exposure of ultramafic bedrock in a series of small waterfalls. Upstream from this area the stream bed contained boulders and cobbles. W This hilly area (31 sq km) has elevations ranging from 91 meters ASL to the highest point reached at Tomah Mountain (329 meters ASL). The thin drift on ridges and poorly 21 CALAlS. ROBBINSTON AR EA CALAlS MAGURREWOCK ‘ RIVER 3‘ ‘ vosE eARiNG 7, . ‘ p0,“, . W L ROBBINSTON EASTERN WESTERN LAKE STREAM A~o STREAM ZmHu . PEMBROK Figure 4: Calais Area Map. 2 COV' spr ced. Dev san: al. Tom. Lit as . org. EVe. bou Cut fau 22 covered, steep bedrock slopes are forested by Northern spruce and fir, with swamp and bog valleys containing mostly cedars and alders. The bedrock consists of intrusive Devonian granite, Ordovician-Cambrian pelites and sandstones, and Ordovician ultramafic volcanics (Osberg et al., 1985). Sediment and water samples were collected from Little Tomah Stream and a tributary (Cabin Creek) that drains Little Tomah Lake (Figure 5). Little Tomah Stream originates as a slow, narrow (1 to 3 meters wide) stream through an organic-rich swamp near the base of Tomah Mountain. Eventually the stream becomes fast-flowing, with a bedrock, boulder, and cobble bed and with numerous waterfalls that cut through bedrock fractures and joints, and over small faults in the granite and ultramafic rock. 23 TOPSFIELO. COOYVILLE AREA ‘Q 1 7) ‘ LiTTLE TOMAH STREAM LITTLE TOMAH LAKE ; (i M's " °' ‘ 3" H, e .. _ .~ (7. g, ..... )6 .~ (0 COOYVILLE TOPSFIELO U I Mole E) Figure 5: Topsfield Area Map. W W For any metal ion in solution, its total concentration will consist of the sum of all aqueous species in that solution, including complexed, colloidal bound, and aquo- complexed ions. Similarly, the total concentration of a heavy metal within sediment will consist of all states in which the metal is found, including interstitial (oxidized and reduced) soluble ions, adsorbed ions on clays, oxides, carbonates, organic matter, and sulfides, and ions within lattice structures of clay and silicate minerals in the nonmobile (residual or detrital) fraction (Gupta and Chen, 1975). The distribution of trace metals among various chemical phases is very important to know in quantifying the behavior of metals in sediment. The manner in which a metal is partitioned among the various phases may reflect the chemical controls operating on adsorption in the water- sediment system. It may also help determine various properties of the metal such as its bioavailability, abundance, and mobility. A quantitative description of trace metal partitioning is one of the objectives of selective chemical extraction techniques. The solid material of sediment can theoretically be divided into distinct phases. A phase is defined in this report as a fraction of the sediment which can release adsorbed trace metals into solution after a change in 24 env usi met par don« ext] res; sele adsc cond meta exch acid exci of ti phaS€ on ex re1Ea Phage anon. OxideE organi 25 environmental conditions. These phases can be extracted by using an appropriate chemical reagent which either degrades a phase or causes a desorption reaction and mobilizes a metal into solution. For the purpose of investigating the partitioning of heavy metals, these chemical extractions are done in a sequence where the weakest-bonded metal ions are extracted first and the strongest-bonded last. In this respect, the extractions are somewhat "selective”, but this selectivity is limited in that the phase which releases adsorbed metal ions may not be a unique substrate but rather a suite of substrates which respond to the changing conditions caused by the reagent. For example, weakly—bonded metals that can be extracted with NH4OAc may be located on exchangeable surface sites of clays, oxides, and humic acids, so the phase or fraction can only be identified as "exchangeable" (Posselt et al. 1968). These extractions can represent an environmental change of the natural conditions that would affect the sediment phases. Exchangeable phases_in sediment will release metals on exchange sites with a change in water ionic composition. garhgnatg_nha§g§ are susceptible to changes in pH and will release associated metals. Reducible fractions are those phases that are thermodynamically unstable under reducing, anoxic conditions. These include easily_redngihle (manganese oxides and amorphic iron hydroxides) and moderately, I:dfl£lhl§.(iron oxides) phages. Oxidizanlg_phg§g§_include organic matter and sulfide compounds that may be degraded 26 under highly oxidizing conditions, which leads to a release of associated trace metals. The strongest-bonded metals in the residual_pha§e§_are not expected to be released under the conditions produced above by extractions over a reasonable amount of time because they are located within crystalline lattices of primary and secondary minerals (Tessier et al., 1979). The partitioning of heavy metals which is determined by sequential selective chemical extractions (SSCE) is identified as operationally defined by the methods used to release metals into solution. The early developments of chemical extractions were made to address the problem of soil fertility and estimates of the availability of trace metals in soils for plants (Jackson, 1958). The non-detrital forms which were extracted included iron oxides, carbonates, plant organic matter, and‘ clay minerals with exchange sites (Table 2). An application of these techniques to pelagic sediments (Chester and Hughes, 1967) used acids, reducing agents such as hydroxylamine hydrochloride, and combined acid-reducing agents to investigate the distribution of heavy metals (Ni and V) adsorbed on sediment particulate surfaces. Iron and manganese oxides are strong scavengers of heavy metal ions, and reducing agents that were developed readily dissolve these phases to release the associated metals. Chao (1972) recognized the importance of developing a separate dissolution of iron and manganese oxides in determining interelemental relationships. Manganese oxides Tabl Chem Cati Hydr‘ Carb; Reduc Basil (Mneo Moder (hydr Oxidi (Orqar Detrit P{Om F 27 Table 2: Chemical phases and extraction methods. Chemical Phase Chemical Extraction Method Cation exchange BaClz; MgClz; NH4OAc Hydrogenous/lithogenous 0.1M HCl; 0.3M HCl; 0.1M HNO3 Carbonates CO treatment; exchange co umns; NaOAc/HOAc-buffer Reducible 1M NHZOH'HCl v/v 25% HOAc Easily reducible 0.1M NH OH°HC1 in 0.01M (Mn-ox and amorphous Fe-ox) HNO @ 5°C; 10% H202 in 0.0 1M HNO ; 0.25M NHZOH'HCI in 0.25M H 1 a 70°C Moderately reducible Oxalate buffer; sodium (hydrous Fe-ox) dithionite with sodium citrate; hydrazine chloride; 0.04M NH OH'HCl in 25% v/v HOAc @ 9g°c Oxidizable 30s H202 @ 95°C, with (organic matter and sulfides) 1N NH4OAc or 0.01M HNO3; fat solvents, e.g. chloro- form, ether, gasoline, benzene, carbon disulfide; KClO3 + HCl-4N HNO3 Detrital silicates EDTA; HF/HClO4; LiBO3 G 1000°C + HF-HNO3 digestion From Forstner and Patchineelam, 1980. 28 differ in solubility from iron oxides in natural environments in response to changes in redox state, charge, and kinds of metals scavenged. Using sediment samples from Maine, Colorado, and Hawaii, Chao (1972) developed a technique to selectively dissolve manganese oxides with little dissolution of any iron oxides. A solution of 0.1 M hydroxylamine hydrochloride prepared in 0.01 M nitric acid (pH = 2) was used. Iron oxides consist of crystalline Fe203, amorphous Fe(OH)3, and some sort of intermediate iron oxyhydroxide. When Mnoz is dissolved, some amorphous Fe(OH)3 is hypothesized to be dissolved as well. For these samples from various states in the USA, the amount of iron was as much as ten times greater than the amount of manganese. Still, using this technique, the manganese oxides were preferentially dissolved. This extraction was also tested in sediments from Maine with large amounts of manganese (0 to 20,000 mg/l) and iron (0 to 500 mg/l) by Chao and Sanzolone (1973). Oxides were extracted in order to determine the concentrations of associated Co, Ni, Cu, Pb, and Zn. The problem being investigated was whether microgram levels of these metals in solution could be determined by the manganese oxide extraction combined with APDC-MIBK chelation and atomic absorption spectrophotometry in solutions that contain large amounts of iron and manganese. Consistent and reproducible results were obtained in their experiments, which is significant since the sediment samples were taken from the Sdll The wit and st: onl par dEtl 29 same areas in Maine sampled in this study (Chao and Theobald, 1976). This technique to distinguish between metals associated with manganese oxides and iron oxides was also used by Chao and Anderson (1974) to look at the behavior of silver in stream sediment from Colorado. They found that silver ion partitioning behavior is significantly controlled by manganese oxides while iron oxides were interpreted to play only a secondary role. Gupta and Chen (1975) used SSCE to study the partitioning of various heavy metals in dredged sediments to determine potential mobile concentrations. In their sequence of extractions, they determined metal concentrations associated with the following sediment components: interstitial water, soluble solid minerals, exchange sites, metal carbonates, easily reducible phases (manganese oxides), organics and sulfides, iron oxides, and lithogeneous (mineral residual) fractions. One interesting aspect of their technique was the use of an oxidizing agent (for organics and sulfides) prior to the use of a moderately reducing agent (for iron oxides). Most studies reverse the order of these reagents, since metals are generally more strongly bonded with organic molecules than iron oxides (Stumm and Morgan, 1981). Gupta and Chen (1975) concluded that if the natural anaerobic conditions were left undisturbed when sediments were dredged, metals in the weakly bonded phases and interstitial water (which are a 30 small fraction of total available metals) would be the most mobile. A significant change in environmental conditions could release metals from the hydromorphic fractions. Chemical extraction methods do not absolutely discriminate between metal concentrations present in the actual soil and sediment phases. However, they can give a general indication of the occupation of binding sites on operationally defined substrates, and they do provide a reproducible technique of analysis (Gatehouse et al., 1977). Even suspended particulates can be analyzed by extractions to determine the modes of metal transportation in streams among soluble, adsorbed, oxide coatings, solid organics, and crystalline minerals (Shuman et al., 1978). Tessier et al. (1980) found that, using SSCE developed in an earlier study (Tessier et al., 1979), most of the natural trace metals in suspended sediment of two southeastern Quebec rivers were in the residual fraction, while iron—manganese oxides and organic matter were important transport phases for bioavailable Co, Pb, Zn, Ni, and Cu. However, man—induced perturbations to the river trace metal input were found to increase relative concentrations in the reducible oxide phases and decrease the importance of the residual fraction as a sink for metals. Robinson (1984) found that by comparing various techniques that extracted metals from exchangeable sites and carbonates, Mn-oxides, organics, and Fe-oxides on particulate coatings (Tessier et al., 1979; Filipek et al., 31 1981; and Robinson, 1982), none were completely specific for releasing metals from a phase. Therefore, they produce some error in determining metal-substrate partitioning relationships. However, it was recognized that the quantitative information which can be learned from this type of study (which distinguishes among substrate phases) is quite important. Forstner and Patchineelam (1980) also pointed out that extractions are often not selective. Repeated treatment of a reagent, for example, may show a further release of adsorbed metals as it attacks stronger phases to some extent. For example, metals can readsorb or precipitate after being released from organics by the action of dilute acids or hydrogen peroxide. Other chemicals have been found to attack different phases after a long extraction time or after a change in pH. One advantage in dissolving acid soluble phases prior to reducible phases is that the reducible extraction will then be more selective, since the buffering capacity of the sediment carbonates was eliminated by the acid. Since particulates in sediment are a diverse mixture of phases (clays, oxides, organic debris), one would expect a wide variety of surface chemistry properties and therefore variable metal adsorption characteristics (Lion et al., 1982). Adsorption would also be expected to vary with any changes in relative proportioning of surface sites among the phases. Since observed experimental metal adsorption 32 behaviors in well defined, single component systems are consistent with conceptual models of adsorption, the use of chemical extractions to determine metal distributions in natural systems would be quite important in the attempt to make predictions of metal behavior in natural systems. Heavy metals are an important group to study in natural aqueous systems because of their sensitivity to variations in environmental conditions, which make them ideal indicators of physicochemical processes (Van Valin and Morse, 1982). SSCE can reproduce potential variations in conditions and isolate how heavy metals behave in response to these changes. Chemical extraction technique efficiency was evaluated by van valin and Morse (1982) through a comparison of metal concentrations determined by extractions with neutron activation and x-ray fluorescence analyses of solid phases. They found that the three methods of analysis produce consistent results, and they concluded that SSCE are a useful method for characterizing the partitioning of trace metals associated with solid phases in sediment. They recommended that the sequential scheme of Tessier et a1. (1979) (with slight modifications) would be very useful. Although the above investigations suggest that SSCE are quite helpful in determining interelemental relationships, there are problems with the selectivity of reagents (which was described above) and readsorption of metals after an extraction. An experiment by Rendell et al. (1980) shows that significant readsorption occurred within sediment after 33 extraction with the reagents for dissolving various sediment phases. This suggests that a misinterpretation of extraction data could occur because the concentration of metals determined in the extraction do not represent the total amount associated with the phases extracted. However, the sediments in their study were extracted for 16 hours and were not applied in sequential order for extracting the weakest to strongest bonded metals. This sequential procedure is very important in identifying quantitative partitioning data and can be tested by conducting steady state tests on the chemical extraction time. In the development of the easily reducible phases extraction, Chao (1977) measured the amount of manganese released versus time of extraction. After a period of 60 minutes, the concentration of manganese in solution leveled off until little manganese was added to solution. This was interpreted to mean that most of the oxide was dissolved. In fact, it was found that over 80% of the available manganese oxide (analyzed after acid digestion) in the sample was dissolved. The study by Chao (1977) is very significant because the extraction techniques were developed using stream sediments collected in the same study areas in Maine that were sampled for this study. Tessier et al. (1979) also measured the effect of extraction time on calcium and iron concentrations in extraction solution for carbonate and moderately reducible phases respectively. Based on these reaction time experiments, a schedule for dissolving the Exc Car Red Oxi Res Ge; the eXC Dha Bag Was dry the ”hi the 34 various phases was developed and is shown in Table 3: Table 3: Tessier et al. (1979) extractions methods. chemical Phase Extraction_methed Exchangeable 1M MgC12 pH=7 at 25°C for 1 hr. Carbonates 1M NaOAc pH=5 at 25°C for 5 hr. Reducible 0.04M NHZOH'HCl in 25% v/v HOAc at 96°C for 6 hr. Oxidizable 30% H202 + 0.02M HNO3 at 85°C for 3 hr., then 3.2M NH4OAc in 20% v/v HNO3 for .5 hr. Residual silicates HF-HClO4 dissolution, then 12N HCl dissolution. Gephart (1982) conducted steady state experiments to find the adequate reaction times for the extraction of exchangeable phases, carbonate phases, easily reducible phases, moderately reducible phases, and oxidizable phases. Based on these experiments, the methodology for this study was formulated (summarized in Figure 6). A comparison test was done to determine whether wet or dry sieving should be done for the sediment samples. Using the stored sample Tom-4, a subsample was dried and sieved while another subsample was wet sieved. The samples were then extracted according to the methods described in Figure 6 (except for the residual fraction) and the concentrations (me Add Ac (1 To 35 SELEEII1E_QHEHIQAL_EKIEA£IIQN§. sediment sample dries (30°C), sieved to (180 micron fraction, 5 gram sample taken for procedure (metals associated with exchange sites such as clays and easily acid soluble such as carbonates) Add 1.0 M NaOAc (pH=5) s 25°C for 5 hours. centrifuge and save extract i (metals associated with Mn-oxides and amorphous Fe-oxides) Add to residue: 0.1 M NHZOH.HC1 v/v 0.01M HNO3 s 25°C for 30 minutes. centrifuge and save extract T (metals associated with Fe-oxides) Add to residue: 0.04M NH 0H.HC1 v/v 25x HOAc s 96°C for hours. centrifuge and save extract erdizah1e_£raciien (metals associated with organic matter and sulfides) Add to residue: 0.02M HNO3 + 302 H302 (pH=2) s 85°C for 5 hours, then 3.2M NH40Ac 9 25 C for 30 minutes. centrifuge and save extract (1 Residual Ecagtion (metals associated with silicates and other chemically resistant minerals) To 0.2 grams of residue: Fusion with Li84 e 1000°C, then dissolution with HN03. Figure 6: Flow chart of chemical extractions for this study. 36 of iron, manganese, chromium, copper, zinc, nickel, and lead were measured on the AAS and converted to ppm whole rock. The percent differences in dry vs. wet sieved samples were calculated for each metal. For the carbonate extraction, the percent differences for all metals had a mean of 7.4% ; for the easily reducible extraction the mean was 14.3%; for the moderately reducible extraction the mean was 15.9t; and for the oxidizable extraction the mean was 34.4‘. For a total of 28 measurements (7 metals, 4 extractions), the mean percent difference was 18%, and there was no consistency for either sieving method being more or less precise. The largest differences were found in the oxidizable extraction, which ranged from 13% difference for chromium and zinc to 77% difference for copper. However, since copper is known to complex with dissolved organics quite readily (Stumm and Morgan, 1981), the wet sieving may have included non-sediment phases (such as interstitial water) which would increase the concentration of copper extracted. Based on these results, the decision was made to dry sieve the sediment samples. hy Th Ni The use acc ful Dur tec W The study areas chosen in Maine (Figure 2) to test the hypothesis were sampled during the summers of 1983 and 1984. The streams with the highest reported concentrations of Zn, Ni, and Pb in sediment were first located on maps of Maine (Post and Hite, 1964; Post et al., 1967; and Nowlan, 1976) The first year (1983) in which samples were collected was used to determine exact locations of streams, their accessibility, and the quality of the sediments in regard to fulfilling the requirements of the research objectives. During the second sampling year (1984), additional field techniques were employed, based on the initial findings. Sediments were collected from oxic zones at the water- sediment interface in the stream beds with a polyurethane scoop, to reduce any metal contamination. water was allowed to run off the sediment sample before transferring it to a one gallon polyethylene bag, which was knotted, double sealed, and transported from the site by backpack. The bags were placed in an ice-packed cooler for storage during the trip back to Michigan, where they were temporarily stored in a walk-in cooler kept at 2°C. Cold storage is required to inhibit bacterial activity which may produce a change in the reactivity of surface sites on sediment particulate surfaces. In 1983, 35 sediment samples were collected in three different areas of Maine: 24 from the Jackman Township 37 fr 531 th( whe bo1 was ane pol rat end bot ana San A11 Wit} Drio labo 38 Area (Figure 3), 3 from the Calais Area (Figure 4), and 8 from the Topsfield Area (Figure 5). In 1984, 30 sediment samples were collected: 16 from the Jackman Township Area (Figure 3), 6 from the Calais Area (Figure 4), and 8 from the Topsfield Area (Figure 5), for a total of 65 samples. In 1984, stream water was also sampled at each site where sediment was collected. First, a 250 ml polypropylene bottle (containing 5 ml of formaldehyde to kill bacteria) was filled with raw water. This was to be used later to analyze the sulfate concentration. Next a 125 ml polypropylene bottle with no preservative was filled with raw water. This would be used to measure alkalinity at the end of the sampling day. Finally, a 500 ml polypropylene bottle was filled, to be used later for the laboratory analysis of concentrations of major cations and anions. This sample of the stream water had to be immediately filtered and acidified with concentrated nitric acid to pH < 1 upon collection, so as not to unnecessarily mix it with the atmosphere. Filtering was done by using a hand operated lever action pump (attached by rubber hose to a vacuum flask) to pull stream water through a 0.45 micron (pore diameter) Millipore filter. The filtered water was then carefully poured into the bottle containing 2-3 ml of acid. All bottles were acid washed with hydrochloric acid, rinsed with double distilled water in the laboratory, and sealed Price to sampling. The 250 ml bottle was prefilled in the laboratory with 5 ml of formaldehyde. The other bottles were :1 Al Hi th Bu th st ca Vi UP th di 8t bo de ac 61 eq' 8&1 Of all the Nat 39 rinsed in stream water at each site prior to being filled. All bottles were kept chilled during and after transport to Michigan. Temperature and pH were also measured at the same time that water samples were collected. Bottles of standard pH Buffer Solution (pH=4 and pH=7) were placed in a bucket of the stream water to allow them to equilibrate to the ambient stream temperature. These buffers were then used to calibrate an Orion Research pH Meter (Model 399A) equipped with a pH electrode (Orion Combination pH, Model 91-05). Upon calibration to temperature and the buffers, the pH of the stream water was recorded by placing the electrode directly into slow streams or into a fresh bucket from fast streams. At the end of the sampling day, the chilled 125 ml bottles were used to calculate alkalinity. This is determined by a titration method using a buret of sulfuric acid solution and a pH meter, as described by Skougstad et al. (1979). A sulfuric acid solution (0.01639 N) with 1 ml equivalent to 1.00 mg CaCO3, is used to titrate a volume of sample to a pH of 4.5, which is the true equivalence point of bicarbonate-carbonic acid under ideal conditions. Total alkalinity is calculated as CaCO3 in mg/l. values for all the streams were quite low (2.2 - 26.6 mg/l CaCO3). All water analysis results are recorded in Appendix 3. 40 Libfllfitfl:¥.flfl£h9dfl. A test was done using one sediment sample to determine if there were any differences in conducting chemical extractions (described in Chapter Three) using wet-sieved or dry-sieved sediments. Wet-sieving involves placing fresh, damp sediment into a clean nylon 180 micron pore diameter sieve, rinsing with double distilled water, and collecting the < 180 micron fraction for drying. Dry-sieving involves drying sediment in clean 1 Liter beakers in a convection oven for 24 to 48 hours at 30°C, and sieving dry sediment with a clean No. 80 Mesh 0.8. Standard Sieve to collect the < 180 micron fraction for storage in waxed containers. Results of the test (described in Chapter Three) suggested that there were few differences between the methods of sieving. Therefore, the sediments were dry-sieved, since this method disturbed the samples the least. The filtered and acidified stream water samples were used to determine the concentrations of major cations: sodium, potassium, calcium, and magnesium. The analyses were completed using a Perkin Elmer Model 560 atomic absorption spectrophotometer (AAS). Results of the major cation analyses are presented in Appendix Three. Major anions are HCO3', 804'2, and Cl'. Bicarbonate is calculated from the alkalinity measurement. Sulfate will be discussed below. Chloride concentrations greater that 10 mg/l are typically analyzed with Mohr's titration. However, since the cation concentrations were very low (Appendix 3), another method fC ch 10 H0 me. all pol Ori det mil vol is Res con tit mil My 61110 had ele Com adde thes 41 for the analysis of chloride was utilized, since the chloride concentrations were expected to also be very low. Chloride concentrations were analyzed using a silver ion electrode. The instruction manual for the Orion Research Model 94-16 electrode describes a titration technique for measuring chloride concentrations less than 5 ppm (10'4 M). A volume of sample is titrated with a dilute solution of silver nitrate while measuring the change in electric potential (in millivolts) with a silver ion electrode and an Orion Model 399A pH meter. The chloride concentration is determined from extrapolating from a linear plot of milliliters of titrant used vs. change in millivolts, on 10% volume-corrected Gran's plot paper. The extrapolated point is used to calculate ppm Cl (1 ml titrant = 0.5 mg Cl). Results are shown in Appendix 3, and the chloride concentrations were quite low. In order to carry out the titration, titrant had to be added to the sample until the millivolt reading was within the range for the blank (distilled water) titration, which was > 290 my. Six of the sixteen samples were much less than 290 mv in spite of the amount of titrant added. It was assumed that these samples had some type of chemical interference which reduced the electrical potential, possibly a dissolved constituent which complexed or neutralized silver ions as quickly as they were added with titration. Therefore, chloride concentrations for these samples could not be measured. 42 Five sediment phases were chosen for this study with SSCE: acid-soluble (including exchange sites on clays, carbonates, and chromium hydroxides), easily reducible phases (manganese oxides and amorphous iron oxides), moderately reducible phases (iron oxides), oxidizable phases (organic matter and sulfides), and residual phases (silicates, lithogeneous minerals). Five gram subsamples were collected from the <180 micron fraction of 61 of the 65 sediment samples (some samples did not have enough fine sediment for extraction). The subsamples were then subjected to the sequential extraction procedures outlined in Figure 7. The residual fraction was determined from a 0.2 gram subsample of the sediment that remained after the oxidizable attack. The 0.2 gram sample was fused with 1.0 gram of lithium metaborate in a graphite crucible at 1000°C for 15 minutes, dissolved with 5 mls concentrated HCl, and then diluted to 100 mls. All extraction solutions were collected in polyethylene bottles and acidified to pH<2. The concentrations of Mn, Fe, Cr, Zn, Cu, Ni, and Pb in solution were determined with a Perkin Elmer 560 atomic absorption spectrophotometer using standards made up with the extraction chemicals to account for matrix affects. In order to carry out estimations of the concentrations of the important sediment phases, it is necessary to determine the amount of total organic carbon (TOC) in the sediment on a percentage basis. A titration technique 43 developed by Gaudette et al., (1974) was used for measuring organic carbon. W The phase concentration factor (PCF) (developed by Forstner and Patchineelam, 1980, and Filipek et al., 1981) was modified by Gephart (1982) in the estimation of phase concentrations and is described below. The amount of each phase must be known or estimated to use a PCF. By making some assumptions about the phases chemically attacked in the sequential selective chemical extractions, the concentrations of the three major phases can be calculated in the following procedure: Q;ganig_mattgr: This is estimated by multiplying the percentage of total organic carbon (TOC) determined analytically in a sample by 10,000 ppm (an approximation for the mean molecular weight of organic molecules) and by 2.2, which is a constant used by Filipek et al. (1981) for organic carbon in biologic material. Manganese_gxides; It is assumed that MnOz and amorphous Fe(OH)3 are dissolved in the easily reducible extraction (Chao, 1972). The weight percent of Mn and Fe in MnOz and Fe(OH)3, respectively is approximately 63.2%. Therefore the concentration of easily reducible phases is estimated by multiplying the sum of Mn and Fe concentrations (in ppm) in the extract solution by 1.6 (100% / 63.2% = 1.6) (Gephart, 1982). th by es th 19 ph ea ca To as be re C01 f0: m0< Pit ter in De: the De: dis Con Get. 44 Lrgn_gxide§; It is assumed that FeOOH is dissolved by the moderately reducible extraction, and Fe is 62% of FeOOH by weight. Therefore, the concentration of this phase is estimated by multiplying the Fe concentration (in ppm) in the extract solution by 1.6 (100% / 62% = 1.6) (Gephart, 1982). These phase concentrations are used to calculate the phase concentration factor (PCF) for a metal partitioned in each of the three major phases. An example of this calculation is shown in Figure 7 for sample Tom-6 from the Topsfield area. The PCF for a metal in a phase can represent the association of that phase with the metal. However, it would be valuable if this PCF could be used to measure the relative association of the phase with that metal in comparison with the other phases. This can be accomplished for a three component system (easily reducible phases, moderately reducible phases, and oxidizable phases) by plotting the relative percentages of these PCF values on a ternary diagram (Gephart, 1982). The concentrations of Cr, Cu, Zn, Ni, and Pb extracted from the three phases were plotted on ternary diagrams as percentages of the amount of metal associated with each of the three phases. PCFs were calculated and plotted as percentages on ternary diagrams to see how metals would be distributed if the phases were present in equal concentrations. In this way some of the variability in the data was reduced. Low concentrations of a substrate will m": CI) Fig 45 Wfiamm A. Calculation of Phase Concentrations i. Oxidizable phase (0X): 2.70 TOC * 10000 * 2.2 = 59400 2. Moderately reducible phase (MR): Fe concentration in leach (1572 ppm) * 1.6 = 2515 3. Easily reducible phase (BR): Mn and Fe concentrations in leach (51.75 + 57.5 ppm) * 1.6 = 174.8 4. Total equals 62090 and therefore the absolute distribution of the phases in percents: QK ER EB 95.7% 4.0% 0.3% B. Nickel concentration in each phase as a percent: Qfi HR ER Ni 1.8 ppm 80% 0.2 ppm 3.9% 0.25 ppm 11.1% C. Calculation of phase concentration factors (PCF) for plotting on ternary diagram (calculations in B / calculations in A.4.): QK £3 £8 PCF 0.84 2.19 39.5 as % 2.0% 5.2% 92.9% The percentage values are plotted on the ternary diagrams. Figure 7: Example calculation of PCF value for sample Tom-6. hav ass dis nor the asso inte stat COHC 46 have a higher PCF value than high concentrations (closer to a value of one) and therefore will have a higher apparent association with a metal, even if the metal is actually distributed equally among the substrates before normalization. This is why the easily reducible phases in the Grand River sediments tend to have such high apparent associations with Zn, Ni, Pb, and Cu (Figure 1). Any interpretations of ternary diagrams must consider this statistical weighting or bias of PCF values on low concentrations. As an example of this statistical bias, consider the values in Figure 7 if the percentages of phases were hypothetically the following: OX 8 95.6%, HR = 2.2%, and ER 3 2.2%. With the percentages of nickel in the phases remaining the same (80%, 8.9%, and 11.1% respectively), the PCF values calculated would be the following: ox = 0.84, HR = 4.05, and ER a 5.05. For plotting on a ternary diagram, the PCF values as a percent would be: OX = 8.5%, HR = 40.7%, and ER 3 50.8%. The actual percentages as calculated in Figure 7 were: OX = 2.0%, HR - 5.2%, and ER = 92.9%. Although the ER phases still have the highest association with nickel, the importance of ER phases in comparison with the HR phases in this hypothetical case would be dramatically reduced. Substrate concentration data were reduced by descriptive techniques to determine values for mean, median, range, and standard deviation. Three areas in Maine were 3am; coul into to d thre sign subs thos. diff+ indi' abunc norms 47 sampled, but it was not clear whether the three data groups could be pooled without introducing additional variability into the data. Therefore, a test of the variation was done to determine if the substrate concentration data of the three areas (Jackman, Topsfield, and Calais) were significantly different. The relative percentages of substrates in the Maine samples were also compared with those of the Grand River to look at substrate abundance differences. The three areas in Maine were also compared individually for substrate abundances to study the effect of abundance variabilities and their influence to the normalization procedures. CHAPTER_£1!E. Walsh The raw metal concentrations for all the chemical phases of the Maine samples are tabulated in Appendix 3. The percentage Phase Concentration Factor (PCF) values for each metal on ternary diagrams will be referred to as the n9;malizgfi_fiat§, The metals Cr, Cu, Ni, Pb, Zn are shown on Figures 8, 9, 10, 11, and 12 respectively. Also plotted on each ternary diagram are the percentages of metal concentrations among the three substrates prior to normalization, which will be referred to as rag_data. The distribution of Cr (raw and normalized data) is shown in Figure 8. The raw Cr data show a distribution mainly between the oxidizable (0X) and moderately reducible phases (HR). Generally the highest percentage of Cr .associated with the easily reducible phases (ER) is less than 15% in most samples. The normalization of data to equal amounts of substrate modifies the distribution so that Cr data clusters near the center of the diagram. However, there is quite a range from high association with the ER phases to high associations with the ox-MR phases. The normalized Cr data cluster location and range (Figure 8) is similar to that observed by Gephart (1982) (Figure 1). This pattern was interpreted by Gephart (1982), Takacs et al. (1983), and Long and Gephart (1982) to indicate that the distribution of Cr may be caused by its ability to exist in two oxidation states ,+3 and +6, in 48 1+9 ER 90 - A .4, A NURMALIZED 70 . . f (D RAW DATA so A A A 30 A 10 00 .. best“; A V V .. a $3099 10 ' 30 7o 90 0X MR CHRDMIUM Figure 8: Ternary diagram for chromium partitioning in Maine sediments. 50 natural environments. The adsorption and therefore relative partitioning behaviors of Cr species would be different than other metal ions with only one dominant valence state. In a stream with a pH in the range of 6 to 7.5, Takacs et a1. (1983) defined a model for the two Cr species: Cr+6 existing 2 and Cr+3 as the thermodynamically stable ion Cr04’ existing as the kinetically stable ion Cr(H20)40H2+1. The negative Cr species can be adsorbed by moderately reducible (MR) phases (such as Fe-oxides), or reduced by organic matter (James and Bartlett, 1983). The positive Cr species can be adsorbed on clays, adsorbed by Fe-oxides, oxidized by 02, and can be adsorbed, oxidized, and desorbed by Mn-oxides. The technique of analysis presented in this study shows a systematic behavior of chromium partitioning, whether the metal comes from a natural source (Maine) or an anthropogenic source (Grand River, Michigan). That is, the normalized data for this study are clustered similarly to the clusters in the Grand River study (Figure 1). The ternary diagram for Cu (Figure 9) shows that the raw data are clustered mostly near the ox phases (> 70%), somewhat with the MR phases (0 to 50%), and very little with the ER phases (< 30%). Normalized Cu data are mostly clustered near the or and ER phase (> 50%) and somewhat clustered near the HR phases (generally < 30%). Normalized Cu data plot in quite a scatter, resembling the plot of Cr data in Figure 8. The distribution of Cu in this study is similar to that of Gephart (1982) (Figure 1) in that the ox 51 90 A A '43 NURMALIZEU &' 70 A. . 0 RAW DATA A A A A A A 50 A 45A.?A A A A A A 'A A 30 A A A A AA A A 0 0A A 0 0 10 A 0 A A9 A 0 u 0‘... A A Jags-.309 .':'a gut "write? :52. LnA o u 9% A \I 0 A! \1 V 10 30 SD 70 0X MR COPPER Figure 9: Ternary diagram for copper partitioning in Maine sediments. 52 and ER phases show the highest association with Cu. However, the MR phases show a higher association with Cu in the Maine sediments than in the Grand River sediments (Figure l). A systematic behavior for Cu is not indicated. The differences in Cu partitioning behavior between the two studies will be discussed below. Raw Ni data in Figure 10 are clustered mostly with the HR and 0X phases, with ER phases showing little association with N1 (< 30%). The normalized Ni data show that the ER and MR phases have higher associations with Ni than the 0X phases (< 20%), which was also found in the results of Gephart (1982) (Figure 1). However, MR phases appear to have a higher association with N1 in the Maine samples than in the Grand River sediment samples. These two studies show a systematic behavior in the observed low association of the 0X phases with N1 (< 20%). I Figure 11 is a plot of raw and normalized Pb data. The raw concentration data show that the MR phases have the highest association with Pb (40 to 90%), the ox phases have a lower association with Pb (< 40%), and ER phases have little association with Pb ((20%). Normalization of the data to equal amounts of phases shows that Pb clusters mostly with the ER and HR phases and is weakly associated with ox phases ((40%). In comparison with the Pb distribution of normalized data in the study by Gephart (1982), the results in this study suggest that the MR phases in Maine sediments have a higher association with Pb than those in Grand River 53 ER 90 AAZAAA .9 A A A NURMALIZED A AAA A 0 RAW DATA 70 AAA AA§A2 ‘2‘. A A 50 :4. A g A 30 A gbm =0“ P O 0 O A Q A Q) A o A 00 9 GD 4* 10 0000 0 o 0 (mime? 09290065 0000 00 @Q 0 Cb A A] \l \l \l 5 __g_eA_\/A V :V 10 30 50 70 90 0X MR NICKEL Figure 10: Ternary diagram for nickel partitioning in Maine sediments. 5h ER 90 A A A NURMALIZED 70 a? (D RAW DATA so A A A 30 A 2 2 O 10 13 V 37: 0X Figure 11: Ternary diagram for lead partitioning in Maine sediments. 55 sediments (Figure 1). The low apparent association of ex phases with Pb is observed to be a systematic behavior for Pb in the results of both studies. The plot of raw Zn data in Figure 12 shows that the MR phases have a high association with Zn compared to the other phases. Normalization of the Zn data shows that Zn clusters mostly with the ER and MR phases (30 to 90%) and is very weakly associated with the 0X phases (< 10%). The apparent association of the ER phases with Zn and the low association of OX phases with Zn show a systematic behavior for Zn in this study and in the results of Gephart (1982) (Figure 1). In the Ni-Pb-Zn partitioning diagrams, where normalized data are generally clustered along the ER and MR phase boundary, Zn data (Figure 12) are less associated with the OX phases than Ni and Pb (Figures 10 and 11 respectively), and Zn data also show a stronger association with the ER phases. This trend for Zn data is similar to that found by Gephart (1982) (Figure 1) and suggests that the technique of analysis depicts a systematic trend for the natural partitioning behavior of Zn. However, normalized Ni and Pb data show a noticeable trend of a stronger association with the MR phases in Maine sediments than in the Grand River sediments. This will be discussed below. It is observed that the partitioning behavior of the five heavy metals have similarities between the Maine sediments and the Grand River sediments from the study by Gephart (1982). For example, the normalized data for each of ER 90 . A NURMALIZED 70 mg 0 RAW DATA A so ., ”m ‘19 30 0 %b%°cgm o o 0W3A “5 c5 0 10 00%&Q98cw0% 1:) V 3V0 V 5V0 V 7‘6 V 9V0 0X MR ZINC Figure 12. Ternary diagram for zinc partitioning in Kaine sediments. 57 the five metals appear to plot in distinct clusters among the three phases. Also, the two major trends found by Gephart (1982) can be seen in the Maine data: 1. clusters of normalized metal data rotate around the ER apex away from the ER-OX phase boundary in the order of Cu < Pb < Ni < Zn, and 2. for the association of Cr with all three phases, Cr data clusters near the center of the ternary diagram. Since there are some differences in the way the normalized metal data cluster on the diagrams in a comparison of these two studies, it is necessary to make a more thorough investigation of the any other differences in the two areas, as well as an investigation of the technique of normalization. WW One factor which may account for the higher associations of Cu, Ni, and Pb with HR phases in Maine than in the Grand River is that two major types of streams were sampled in Maine; fast flowing and slow flowing, boggy streams. The distinction was in the amount aeration of stream waters observed in the field. The rates of flow in these stream types may cause a change in the oxidizing- reducing environment. Although the values for redox and dissolved oxygen were not measured for the streams, the fast flowing streams would typically be more oxidizing while the slow flowing streams would tend to be reducing. The different redox conditions may cause a difference in the adsorption behaviors of metals. 58 The metal partitioning data for these two stream types were divided into two groups, Fast and Slow, but only 51 of the 61 sampling sites could be positively identified as being fast or slow flowing (some dry beds were sampled), with 26 in the Fast group and 25 in the Slow group. Figure 13 summarizes the distribution of metals between the two groups. No differences were detected in how partitioning data for Cu, Ni, and Pb were distributed in fast and slow streams, so it was necessary to consider some other explanation for the differences in adsorption behaviors of copper, nickel, and lead between the two studies. It was found, however, that the Cr data in the slow streams seems to plot in two distinct groups. This may be caused by the fact that Cr exists in two valence states, Cr+3 and Cr+6. If there was a factor in the slow streams which favored the existence of different adsorption behaviors of these two species, the differentiation of these behaviors may be identified. lo 30 so 70 so 0X HR NI SLOW-FLOR ER O u u n n u- :1 [AA 8 u no 30 $5 70 so PB SLOW-FLO! 59 no u u n ”A M n n no 30 so 10 é: 0X MR CU FAST-FLDH ER TB"§a“§a"7“o"Eo ox MR NI FAST-FLOW ER ‘ PB FAST-FLOW Figure 13: comparison of metal partitioning in fast-flowing and slow-flowing streams. ZN SLDV-FLOV ER in 30 50 7D 90 0X MR CR SLOW-FLOW Figure 13: ZN FAST-FLOW ER \l I AIAA“-I 73 +30 "'"soffi 33+ 1;: 0X MR CR FAST-FLOW continued. 61 Wu: Since three distinct areas in Maine were sampled (Figure 2), with different bedrock types and mineralizations, it was hypothesized that the normalized partitioning data may plot distinctly as well. This may explain some of the variability in the data. The data were divided into the three sampling areas (Jackman, Calais, and Topsfield, Figure 2) and plotted on separate ternary diagrams. All the metals were studied in this manner in order to see if partitioning trends are similar among the three sampling areas. mm The three ternary diagrams for chromium (Figure 14) show that data from the Calais and Topsfield areas cluster either near the ER phases or only between the ox and HR phases, but that data from the Jackman area clusters almost equally among the three phases. One explanation for the data in the Calais and Topsfield areas plotting at such extremes is the low detectability of chromium by the methods used for this study (Flame AAS). For all samples, chromium was in low concentration in the extractant solutions. Therefore, since the normalization procedure is biased toward the phases with the lowest concentrations (ER), any concentration of chromium present would plot mostly with that phase. If that extractant solution had no detectable chromium, then the chromium would be distributed among the other two phases 62 .ocdw: ed nmoun oounu any new nEmHawdc auscuou E=«Eouso ”ca onsmau cu cam—ammo» cu z> .05, where P represents the 95% confidence level). The ANOVA for the MR phases (Table 8) was interpreted to show that these means are significantly different (P < 0.05). The ANOVA for the ER phases (Table 9) shows that the means are significantly different (P << 73 Table 5: Summary statistics for raw phase concentrations and log phase concentrations in Maine sediments. W W Sample size 61 61 61 Mean 4634 14076 31431 Median 819 13408 23100 variance 7.66E7 7.36E7 7.30E8 Stan. dev. 8750 8577 27026 Stan. error 1120 1098 3460 Minimum 118 2515 2420 Maximum 35128 41290 149600 Range 35010 38765 147180 Mines. WW Sample size 61 61 61 Mean 3.12 4.06 4.38 Median 2.91 4.13 4.36 variance 0.43 0.08 0.11 Stan. dev. 0.65 0.29 0.33 Stan. error 0.08 0.04 0.04 Minimum 2.07 3.40 3.38 Maximum 4.55 4.61 5.17 Range 2.47 1.21 1.79 Table 6: Summary statistics for the raw phase concentrations (ppm) for each of the three areas sampled in Maine. -ER_nhases Calais_Area Sample size Mean Median variance Stan. Dev. Stan. Error Minimum Maximum Range 9 1085.2 607.2 1.1936 1090.7 363.6 283.2 3170 2886.8 ER_nhase§ Tensf1e1d_Area Sample size Mean Median variance Stan. Dev. Stan. Error Minimum Maximum Range 13 498.3 463.2 62298.7 249.6 69.2 174.8 1025.6 850.8 EB_nhases lackman_5rea Sample Size Mean Median Variance Stan. Dev. Stan. Error Minimum Maximum Range 39 6831.1 1783.6 1.0788 10335.5 1655.0 118 35128 35010 MB_nhases Qx_nhases. 9 9 15111.1 25691.1 13760 19360 5.61E7 1.53E8 7492.9 12363.1 2497.6 4127.7 5760 15620 32416 49280 26656 33660 MR_nbases__________Qx_nhases 13 13 6606.2 50560.6 7326 36250 3.9237 2.1439 6257.1 46296.3 1735.4 12640.6 2515.2 2420 25964 149600 23466.6 147160 153.236395 0x_nhases 39 39 15594.2 26332.3 15584 22660 7.9937 2.6336 6936.3 16616.1 1431.0 2692.7 2764.8 2660 41260 63600 36515.2 60740 75 Table 7: ANOVA for Maine log oxidizable phase concentrations. MW Number of Groups: 3 (Jackman, Calais, and Topsfield Areas) Confidence Level: 95 Whoa . Source of Sum of d.f. Mean Square F-ratio Sig. Wm Level Between groups 0.18623 2 0.09312 0.850 0.4327 E11n1n_g;gnp§ 6.35527 58 0.10957 11 Total 6.54150 60 Table 8: ANOVA for Maine log moderately reducible phase concentrations. Maintenance“ Source of Sum of d.f. Mean Square F-ratio Sig. yariatinn______finnares Lgygl Between groups 0.64273 2 0.32136 4.375 0.0170 flithin_groups 4.25997 58 0.07345 Total » 4.90270 60 Table 9: ANOVA for Maine log easily reducible phase concentrations. Wm . Source of Sum of d.f. Mean Square F-ratio Sig. Rhianna—Mam Lml Between groups 5.36605 2 2.68302 7.679 0.0011 W521 58 0.34940 - Total 25.63126 60 76 0.05). This may be explained by the fact that there were a number of very high concentrations of ER and MR phases in the Jackman Township Area, which was described above. Since the phase concentrations in Maine sediments are only approximately lognormally distributed, the nonparametric Kruskal-wallis analysis was also done to compare the three areas in Maine. In this analysis, the data are ranked from smallest to largest, the ranks are summed for each sample, and a test statistic value is calculated and compared with tabulated values from a Chi-squared distribution. The results of this analysis for the three phases are summarized in Table 10. These results are interpreted to show that the means for the 0x phases from the three areas are not significantly different (P > 0.05) while the means for the ER and MR phases are significantly different (P < 0.05). These results are similar to those of the ANOVA. Therefore, it has been found that there are statistically significant differences in substrate concentrations between the three areas sampled in Maine, suggesting that three distinct p0pulations were sampled. This may be a cause for the variance seen between the two studies. To address question 2, the absolute (raw) and relative (percentage) concentrations of the substrates will be compared for the Maine and Grand River studies, and compared among the three sampling areas in Maine. The phase 77 Table 10: Kruskal-wallis analyses for log phase concentrations in Maine sediments. Number of Levels: 3 (Jackman, Calais, and Topsfield Areas) Confidence Level: 95 W Mania—Wank Jackman 39 29.2949 Calais 9 28.5000 Igngfield 13 37-8462 . Test statistic = 2.47207 Significance level = 0.2905 WWW: Wank Jackman 39 34.3462 Calais 9 34.2222 Igngfiigld 13 18.7308 Test Statistic = 7.89209 Significance level = 0.019331 MW Lani—SamlLflZL—Axemmnk Jackman 39 36.8718 Calais 9 25.2222 xgnafiigld 13 117-3846 Test statistic = 12.8662 Significance level = 1.60743E-3 78 concentration values in both studies have approximately lognormal distributions. The medians rather than the means will be used as measures of central tendency (Sokal and Rohlf, 1969) to compare substrate concentration data, because a few very high concentration values tend to bias the mean values (Sokal and Rohlf, 1969). Also, the larger values for phase concentrations have been suggested to influence metal partitioning in the normalization procedure. In the Maine sediments, the median absolute concentrations of the three phases were almost twice as large as the concentrations found in the Grand River sediments (Table 11). However, the relative concentrations of phases (calculated as percentages) for the two study areas are very similar: Maine Grand_Rixer 33 = 2.2% ER = 2.1% MR = 35.96 x3 = 38.6% 0x = 61.9% ox = 59.3% Since the partitioning behaviors of the metals among the adsorbing phases between the two study areas are very similar, and the relative abundances of the phases are also quite similar, this suggests that the relative abundances of a phase are more important than the absolute abundances in controlling metal partitioning. To address question 3, the process of normalization will be examined by looking at how metal partitioning behaviors in the three areas in Maine are affected by 79 Table 11: Comparison of sediment phase concentrations (ppm) for Maine streams and the Grand River, Michigan. Median_¥alnes. Maine Grand_nixer Ehass_Ixne ConC- 48:11.11 Conc. Bell_1 Easily Reducible 619.0 2.2 365.0 2.1 Mod. Reducible 13406.0 35.9 6592.0 36.6 Oxidizable 23100.0 61.9 10120.0 59.3 80 normalizing the data. The absolute and relative abundances of substrate concentrations for each of the three sampling areas in Maine are shown in Table 12. The absolute abundances for the substrates from the Topsfield area are quite different from the Calais and Jackman substrate concentrations. The relative percentages for the substrates in the Topsfield area are also different from the Calais and Jackman areas (SER and SMR in Topsfield are lower than in the other two areas, and 80X in Topsfield is higher than the other two areas). Although the relative percentages for substrate concentrations in Topsfield differ from the other two areas, the partitioning data for the Topsfield and Calais areas have been shown to cluster similarly on the ternary diagrams. Therefore, it in concluded that in spite of substrate concentration differences among sampling areas, there are still similarities in metal partitioning behaviors. variability in substrate or metal concentrations is reduced by normalization, but very low substrate concentrations will have higher apparent associations with metals simply because of the calculation technique. However, normalization does not appear to influence the partitioning of the data on ternary diagrams, even if absolute and relative concentrations of substrates vary. When the Maine results for Cu partitioning are compared with the Grand River results, Cu has a higher association 81 Table 12: Comparison of sediment phase concentrations (ppm) for three sampling areas in Maine. Wines. Calais Area Phase Tyne Cone. 3:1. 3 Eas. Reducible(ER) 607.2 2 Mod. Reducible(MR) 13670 41 Oxidizable(OX) 19360 57 Topsfield Area ER 463.2 1 MR 7328 17 0X 36250 82 Jackman Area ER 1783.6 4 MR 15584 39 0x 22660 57 82 with MR phases in Maine sediments. This is not explained by the substrate differences and the discussion above which addressed the three questions. Since it was found that large concentrations of substrates can affect normalization and also the partitioning behaviors depicted on ternary diagrams, it was hypothesized that perhaps large metal concentrations would also affect metal partitioning. In Appendix 3, which has a tabulated listing of the raw metal concentrations found for each of the five chemical extractions, all samples with very high concentrations of Cu ’ in the hydromorphic fraction (> 100 ppm whole rock) were isolated and are listed in Table 13. This concentration of Cu was chosen because it was equal to the mean of the highest concentrations of Cu that were found in the residual phases (i.e. Cu with an origin in mineralization) from all areas sampled in Maine. 0f the 17 samples with high Cu concentrations, 15 (88%) of these were found to be highly associated with MR phases on the ternary diagram. However, these 15 samples represent only 38% (15 out of 40) of the total number of samples which show the high association of Cu with the MR phases. This leaves 40 of the 61 samples (66%) which have a high association with MR phases, but these 40 samples do not have a high Cu concentration, so high Cu concentrations do not explain all of the variability. The difference in Cu behavior between Maine and the Grand River studies may be caused by the source of the metal, since Cu in the Grand 83 Table 13: List of samples with high concentrations of copper (> 100 ppm) in the hydromorphic phases. Cal-2 Tom—1 Jac-14 Cal-6 Tom-3 Jae-16 Tom-4 Jae-17 Jac-19 Jac-Zl Jae-22 Jac-23 Jac-27 Jae-28 Jae-31 Jae-33 Jac-35 84 River is more likely to have come from an anthropogenic source while Cu came from a natural source of weathered rock (especially mineralized zones) in Maine. However, the effects of mineralization were not explored in this study. The literature suggests that an association of Cu with MR phases (especially Fe-oxides) is not uncommon in experimental and field studies (to be discussed below). The following findings were made from the comparison of the Grand River and Maine substrate data: I. The substrate concentration data from the Grand River and Maine study areas are approximately lognormally distributed. 2. Both analysis of variance and Kruskal—wallis analysis show that the OX phase concentrations for the three areas sampled in Maine are not significantly different, and the concentrations for the ER and MR phases are significantly different. 3. In a comparison of the substrate concentrations of the Gephart (1982) study and this study, the absolute concentrations are different, but the relative concentrations are quite similar. 4. In a comparison of the substrate concentrations among the three areas sampled in Maine, the absolute and relative abundances of substrates in one area (Topsfield) are different from the other areas (Calais and Jackman). However, metal partitioning behaviors in sediments of Topsfield and Calais are very similar. 85 5. The analyses of variance and Kruskal-wallis analyses are interpreted to mean that the ER and MR phase concentrations are significantly different among the three areas studied in Maine. This could be caused by some very large concentrations of these phases which are found in the Jackman Township Area. 6. The normalization technique is not weighted on the ER phase concentrations in Maine sediments, as it is in the study by Gephart (1982). In some samples, ER phases made up a large fraction of the sediment in Maine. This could explain why there is an even clustering of Ni, Pb, and Zn between the ER and MR phases on the ternary diagrams (Figures 10, 11, and 12 respectively). This is different than the results of Gephart (1982), where these metals are mostly distributed among the ER phases. In conclusion; 1. there are significant differences in ER and MR substrate concentrations which may cause the variability in the results of the Maine and the Grand River studies; 2. the absolute abundances of substrates are not as important as relative abundances in controlling metal partitioning; and 3. the process of normalization is able to take substrate concentration differences into account in the 86 construction of ternary diagrams to depict partitioning behaviors. MW The partitioning behavior of chromium has been discussed by Gephart (1982). Cr data is mainly associated with the MR phases in the raw data, but normalized data show a significant association of Cr with the ER phases. This is supported by other studies of Cr in natural sediments (Moore et a1, 1984) where the adsorption of the Cr+6 species by iron oxides is dominant in natural waters with a pH between 6 and 8, while at the same time the adsorption of the Cr+3 species by manganese oxides and subsequent oxidation to Cr+6 can also take place (Leckie et al., 1980). Copper, zinc, lead, and nickel are adsorbed by Mn- oxides in the ER fraction. Using nine synthetic manganese oxides, McKenzie (1980) found that all of these oxides have a higher association with lead than with copper, zinc, and nickel. This was suggested as the reason that lead is found accumulated in the manganese oxides of soils. The results of lead adsorption on natural Mn-oxides in Maine and the Grand River differ from the results of McKenzie (1980) with lead not as highly associated with Mn-oxides as copper, zinc, and nickel. However, the use of synthetic manganese oxides in laboratory adsorption experiments may not be representative of the adsorption of metals in many natural systems, since there are many other factors which must be considered. Catts and Langmuir (1986) recognized that the application of the 87 results of adsorption experiments with synthetic oxides to adsorption on natural manganese oxides is somewhat tenuous and cannot predict the magnitude of changes in metal adsorption accurately. For the study of stream pattitalataa_and gaatings by Filipek et al., 1981 in the vicinity of a polymetallic sulfide deposit, pattiaalata organics also were associated with the highest absolute concentrations of copper and zinc (as found by McKenzie, 1980), while lead was mostly associated with Mn-oxides. However, a competition analysis and normalization technique determined the following sequences for the relative importance of the association of gaathg phases with a metal: for Cu, organics > Fe-ox > Mn- ox, for Zn, Mn-ox s organics > Fe-ox, and for Pb, Fe-ox > organics > Mn—ox. These differed from the partitioning of metals in the pattitalataa and from the results from Maine and the Grand River. However, if the phases extracted in the Maine and Grand River studies are considered a mixture of caatinga,and pattitalataa in the sediment, the partitioning results that are found in these studies are similar to the results of the study by Filipek et al. (1981) which are shown above. In the results from Maine, the high association of MR phases (including Fe-oxides) with copper was different than the copper distribution found by Gephart (1982) in the Grand River sediments. However, an association of copper with Fe- oxides after normalization to equal amounts of phases has 88 been noted in a number of other studies. Filipek and Owen (1979) found that the moderately reducible fraction of Little Traverse Bay sediments showed a moderate association with copper, containing an average of 24% of the non- lithogeneous copper. In stream sediment near a porphyry copper deposit in Arizona, 56% of the total normalized copper among the Mn-oxides, Fe-oxides, and oxidizable phases was found associated with the Fe-oxides (Filipek and Theobald, 1981). The nan;natmalizad,distributions of metals that were found in Maine show a similarity to the adsorption characteristics of estuarine particulate matter studied by Lion et al. (1982). In that study, 65% of the copper was associated with the oxidizable phases (compared with 72% for Maine sediments) and 70% of the lead was associated with the reducible phases (compared with 76% for Maine sediments). The same types of chemical extractions were used in each of these studies. Forstner and Patchineelam (1980) looked at the normalized partitioning of metals in the polluted sediments of the Rhine River and found lead, copper, and chromium mostly associated with MR fraction, zinc and copper associated with the ER fraction, and copper and lead associated with the OX fraction. These results are very similar to the normalized results from the sediments in Maine as well as those of Gephart (1982). 89 The raw data distributions of the 5 metals studied in Maine are also similar to the distributions found by Tessier et al. (1980) in the Yamaska and St. Francois Rivers of Quebec. Among hydromorphic fractions (exchangeable, carbonates, Fe-Mn oxides, and organics) non-normalized copper was mostly associated with organics (60%) and Fe-Mn oxides (23%), nickel was mostly associated with the Fe-Mn oxides (65%), lead was associated mostly with Fe-Mn oxides (56%) and organics (23%), and zinc was associated mostly with the Fe-Mn oxides (59%) and very little with the organics (8%). In the Maine hydromorphic fractions, the percent associations of non-normalized metal data are: copper with organics (72%) and Fe-Mn oxides (28%), nickel with Fe-Mn oxides (60%), lead with Fe-Mn oxides (80%) and organics (20%), and zinc with Fe-Mn oxides (86%) and organics (145). Similar results were found for another stream in Quebec which was located near a rock zone with Zn- Cu-Pb mineralization (Tessier et al., 1982). Moore et a1. (1984) have also found similar partitionings for non-normalized data for copper, zinc, nickel, and lead, especially result of the low concentrations of lead, zinc, and nickel associated with organic matter (5-10% for lead, 10% for nickel, and < 5% for Zn). Chromium was distributed among exchangeable (1.1%), easily reducible (2.7%), oxidizable (28.3%), and moderately reducible (67.9%) phases. These are similar to the percentages found in Maine for chromium. Wk Conclusions. The impacts of particle-bound metals on bioavailability in natural systems are difficult to predict because of the differences that occur when metals are bound to different binding sites on particle phases. However, quantitative methods for describing metal partitioning among binding sites in natural sediments have not yet been developed. Quantitative descriptions in the past have involved the use of selective chemical extractions which remove heavy metals from sediment phases. At equilibrium, the partitioning of a heavy metal ion among adsorbing substrates at the water-sediment interface is believed to be influenced or controlled by: l) the binding capacity of each substrate or operationally defined phase; 2) the binding intensity of each metal ion to each phase; 3) the absolute abundance of the adsorbing phases in the sediment; and 4) parameters such as the pH, pe, temperature, and the concentrations of major cations and anions (Luoma and Davis, 1983). The development of quantitative partitioning models will require the use of some assumptions regarding the binding characteristics of adsorbing phases, with concentrations of these phases measured by operational techniques (chemical extractions). Presently these techniques for studying metal partitioning are the only 90 91 suitable methods that provide consistent results in multicomponent, natural systems. A major objective of this type of research is to quantitatively define natural water-sediment systems, in order to make predictions on how a metals in these systems will behave under certain environmental conditions. Attempts to accomplish this have involved measurements of as many conditions as possible in the natural system, and have reduced the amount of variability to the extent that the important controls on the system can be identified. The next step is to predict how the behavior of the specific component may be altered when these controls are changed. An important portion of the study of an aqueous system is the partitioning of the components adsorbed on surface sites at the water-sediment interface. This has been the focus of my study, and is important because systematic trends in natural metal distributions in sediments can be depicted with these methods and are consistent with metal behaviors determined from studies of other natural systems and from experimentation. This may help lead to the development of quantitative models of metal adsorption in natural systems. The conclusions for this study are: 1. Using chemical extractions and the normalization of metal concentrations to equal amounts of adsorbing phases, distinctive cluster patterns for metal partitioning behaviors in natural systems can be depicted on ternary diagrams. 92 2. The additions of anthropogenic versus natural sources of chromium, copper, nickel, lead, and zinc show similar . clustering patterns, and these partitioning behaviors are consistent with the results of other studies on adsorption behaviors for these metals. 3. An evaluation of the normalization of data to equal amounts of substrates has shown that large ranges of values in substrate concentrations can cause a bias towards the least abundant component in the depiction of metal partitioning on a ternary diagram. However, the patterns of metal behavior can still be interpreted in light of adsorption theory and experimentation. 4. A comparison of the absolute and relative concentrations of adsorbing phases in the Grand River sediments with the three sediment sampling areas in Maine has shown that the absolute abundances of an adsorbing phase are apparently not as important as the relative abundances in controlling metal partitioning. 5. Although distinct metal behaviors are indicated on the ternary diagrams, normalized data on the diagrams do not quantify the data sufficiently to be used alone in predicting metal behaviors in various water-sediment environments. The effect of other quantifiable factors on metal partitioning behaviors depicted on ternary diagrams should be analyzed. APPENDIX 1: Geochemistry of Heavy Metals AEEENQLKAL Wm Our understanding of the behavior of heavy metals in natural aqueous systems will be very important as we increase industrial metal additions to our environment. The behavior of metals in adsorption/desorption reactions are a function of many variables, including their basic chemistry in natural geologic systems. Therefore, it is important to evaluate geochemical characteristics of metals in order to adequately model and predict adsorption behaviors. Metal atoms are characterized by their tendency to form positive ions (Murray and Dawson, 1980). As a class of solid substances (i.e. "metal"), they have properties of high electrical and thermal conductance, reflectivity, and mechanical strength and ductility (Cotton and Wilkinson, 1980). One exception to this is metallic mercury, which is a liquid at standard temperature and pressure. Most metals exist in the form of compounds and minerals. There are four main categories of metal ores: 1. Highly electropositive metals (Group IA: Li, Na, K, Rb, Cs; and Group IIA: Be, Mg, Ca, Sr, Ba) are primarily found as salts such as halides, sulfates, nitrates, and carbonates. Group IIA elements are most abundant as sulfates and carbonates. 2. Aluminum and the more electropositive transition elements (Sc, Ti, v, Cr, Mn, Y, Zr, Nb, La) form naturally-occurring oxides. 3. Other transition elements (Co, Ni, Cu, Zn, Cd, Hg), Group IIIA elements (Ga, In, T1), and Group IVA 93 94 elements (Ge, Sn, Pb) are usually found as sulfide minerals. 4. Relatively unreactive metals (Ru, Rh, Pd, Ag, Os, Ir, Pt, Au) occur naturally as free metals or in a compound (such as a sulfide, AgS, or substituting for another element in a mineral, such as Rh-rich molybdenite) (Murray and Dawson, 1980). Goldschmidt (1954) formed a set of rules to describe the distribution of elements in the earth's crust. Primary distribution was interpreted to show that most elements can be classified in the following groups: stfiataphtla if they are relatively inert and commonly exist with native iron (includes Group VIII elements Fe, Co, Ni, etc and Mo, Ge, Sn), lithanntla if they characteristically are found concentrated as silicates (Groups IA, IIA, 1118, and IVB including Li, Na, K, Rb, Cs, Be, Mg, Ca, Sr, Ba, Sc, Y, Ti, Zr, Hf, V, Nb, Ta, Cr, W, Mn; also 0, halides and rare-earth elements), atmapntla if found as natural gases (for example H, N, C, O, halides, and inert gases), and biaahtla if they are enriched in organisms (includes many elements common in all the other groups) (Goldschmidt, 1954; Krauskopf, 1979). For the secondary distribution of elements (especially heavy metals, which for this paper are considered to be the first row transition elements), Goldschmidt (1954) believed that the behavior of an element in the formation of crystal lattices was controlled by their ionic radius. Three general rules were empirically deduced: 1. If two ions have the same radius and charge, they will be incorporated into a crystal 95 lattice with equal facility. 2. If two ions have the same charge and similar radii, the smaller ion will be preferentially incorporated and will form a stronger bond. 3. If two ions have similar radii but different electrical charge, the more positively charged ion will be preferred and will make the stronger bond (Burns, 1970; Mason and Moore, 1982). Although these principles helped to explain some of the trends in trace metal distributions in rocks and minerals, there were several exceptions to Goldschmidt's Rules, such as the behavior of zinc in ferromagnesian silicates (Burns, 1970). Although zinc and iron have identical charge and ionic radii, zinc can substitute for either iron or I magnesium (magnesium has a smaller ionic radius) (Wedepohl, 1969). Ringwood (1955) modified these rules with the deduction that the relative bond strengths of ions in crystal lattices can be based on their respective electronegativity. If two ions have similar ionic radii and electric charge, the ion with the lower electronegativity value will be preferentially incorporated into the lattice and will form a stronger, more ionic bond than the other ion (Burns, 1970). These rules by Goldschmidt (1954) and Ringwood (1955) lack generality in interpreting trace element distributions, particularly with the heavy metals. For example, the geochemical distribution of nickel in magmatic crystallization is unusual and cannot be explained by its 96 ionic radius and electronegativity alone. Nickel would be expected to behave like magnesium since it has essentially the same radius and charge. However, nickel behaves as if its effective radius were less than that of magnesium, which can be observed by comparing unit cell lengths of NiZSiO4 (281 Angstroms) and MgZSiO4 (292 Angstroms) (Mason, 1966). For reasons such as this, other chemical properties have been applied to more accurately describe metal distribution behaviors. The crystal field theory (CFT) of chemical bonding in transition metals has been used to help explain heavy metal behaviors that are exceptions to the above rules. CFT assumes that the only interactions between metal ions and ligands are electrostatic ones (Huheey, 1972). The angular distribution of electron density about a nucleus in terms of wave functions can best be represented by molecular orbitals. The partial filling of the 3d orbitals with electrons characterize the first row transition elements. Induced magnetic and electric fields by coordinating ligands can cause an energy separation of the normally degenerate orbitals. In this case, ions with the same charge can each have different crystal field stabilization energy (CFSE) and therefore may form different types of complexes. Differing energy environments cause a unique energy separation (or crystal field splitting) for each transition metal. The magnitude of the splitting is dependent on: 1. valence of 97 the ion. 2. Nature and type of coordinating ligands. 3. Interatomic distance. 4. Symmetry of the coordinating ligands (Burns, 1970; Huheey, 1972). Evidence for the importance of CFSE can be seen in making predictions of the lattice energies for heavy metals. Predictions were good for Mn+2 and Zn+2, for example, but discrepancies for the metals Cr+2, Fe+2, Co+2, Ni+2, and Cu+2 were all explained by CFT. CFT also explains unusually stable aqueous complexes that have been found. Although aqueous Co+3 is thermodynamically unstable and is easily reduced to Co+2 by water, if certain ligands are present in an aqueous solution, Co+3 is a perfectly stable ion (Huheey, .1972). In summary, the distributions of heavy metals in geochemical systems are related to ionic size, charge, and bond character. The unique d-orbital bonding properties of transition elements facilitates the interpretation of their geochemical behavior in rock and sedimentary environments. However, metal behaviors are also controlled by the physical and chemical properties of the environment. For example, in an aqueous system, some of the important conditions to consider include the system's oxidation-reduction potential (pe), pH, temperature, solution-mineral equilibria, aqueous complexation, aqueous ion speciation, and elemental and hydromorphic adsorbent concentrations (Stumm and Morgan, 1981). The prediction of the behavior of a metal in any system is dependent on the quantification of these factors. APPENDIX 2: Adsorption Theory and Experimentation AEEEHDLX_Z Winn One important objective of this study is to interpret the metal adsorption behaviors which are found using the techniques of Gephart (1982) in light of current metal adsorption theory and the results of laboratory experimentation. By doing this I will assess the ability of these techniques of chemical extractions and normalization of partitioning data to describe metal partitioning in natural systems. This is a necessary step if we are able to develop models for predicting metal adsorption in multi- substrate systems. WW At a liquid-solid interface (or water-sediment interface in this study), adsorption can be described as an attachment of solute ions to the surface of the solid. Although there is no generally accepted model for adsorption that explains all observed surface phenomena, several models have been developed and refined that can account for many experimental observations. Two accepted principles in adsorption theory are that the solid-solution interface has an associated surface charge and an electric potential gradient that extends from the interface into the liquid (or solution) (Leckie et al., 1984). Adsorption can be referred to as non-specific (electrostatic bonding) or specific (caused by London-van der waals forces, covalent bonding, hydrogen bridges and bonding, hydrophobic bonding, steric 98 99 effects, or specific ion effects that form a charged surface). The surface charge on a solid which will be a potential binding site for dissolved ions can originate in three ways: 1. Chemical reactions which involve the surface and ionizable functional groups on dissolved species. Some of these functional groups are -OH, -COOH, and -OPO3H2. The charge that results is strongly dependent on the pH of the solution. A more basic pH will tend to form a negatively charged surface. A more acidic pH will form a positively charged surface. 2. Lattice imperfections at the solid surface can result in isomorphic replacement of atoms by similarly sized ions with a different charge. This will cause a charge imbalance which can be satisfied by the attraction of charged aqueous ions to the surface site. For example, Al+3 could replace Si+4 in an array of 8102 tetrahedra and result in a positive charge deficiency and potential binding site. 3. The adsorption of a surfactant ion can produce a surface charge. These ions may be bound by London-van der Vaals force interactions and hydrogen or hydrophobic bonding (nonspecific adsorption), such as with some organic surfactant ions. The binding site available to charged solute ions is made up of the charged, adsorbed surfactant ion bound to the solid surface. These surface charge sites are examples of nonspecific adsorption. The electrical potential gradient for an interface is usually defined by a pair of ions that are present on the 100 surface and in solution, and are called potential- determining ions (PDIs). On an oxide surface, for example, H+ and OH’ are uSually chosen for the PDIs. Each system has a condition at which both the P013 (at unique concentrations) are equally adsorbed, known as the zero point of charge (ZPC). For an ideal system, the surface charge and surface electric potential gradient are zero at the ZPC. ZPC is expressed as a pH for solid phase oxides. Ideally, when solution pH is greater than the oxide ZPC, the oxide surface will have a negative charge, and when solution pH is less than the oxide ZPC, the surface will have a positive charge. Ions of opposite charge in solution will be attracted to the charged surface (Leckie et al., 1984). The distribution of charges, ions in solution, and an ' electrical potential at the solid-solution interface make up an electric double layer (EDL). One layer is the fixed or surface charge of the solid and the other layer is the diffuse distribution of charged ions in solution. This is referred to as the Guoy-Chapman diffuse charge model, and considers only electrical interactions. Charged ions in the diffuse layer can be counterions (opposite charge of the surface) or co-ions (same charge as surface). Stern (Grahame, 1947) refined this theory to account for nonelectrostatic adsorption. Specifically adsorbed ions are closest to the fixed charge layer (or they account for the charge themselves) and form the "Stern Layer". Ions in the Stern Layer can be subjected to electrostatic and/or 101 specific interactions. Ions in the more diffuse, bulk solution extending from the surface make up the Guoy Layer. This was based on the hypothesis that an ion retains its hydration sphere during adsorption. Grahame (1947) refined the theory even further by suggesting that only specifically adsorbed ions (resulting from non-electrostatic interactions) can lose their hydration spheres in their approach to the surface. Together, these are referred to as the Guoy-Chapman-Stern-Grahame (GCSG) adsorption model, although it is still considered an oversimplification of surface phenomena (Leckie et al., 1984). Further developments of this model by other workers have explained observed anomalies (such as the unusually large surface charge of metal oxide/solution interfaces), and the assumption of interfacial ion pair formation by adsorbed ions has led to the most refined model to date, known as the Stanford Generalized Model for Adsorption (SGMA) (Leckie et al., 1984). Another method to model adsorption is by studying the thermodynamics of adsorption reactions (James and Healy, 1972). The total free energy of adsorption ( AiGads) was hypothesized to be the sum of three components: electrostatic work (.Achoul), total specific adsorption energy ([3 Gchem’! and the change in secondary solvation energy (A Gsolv) . W11 The adsorption of heavy metals by organic matter and hydrous metal oxides in sediments is usually described in 102 the context of the above theories. Hydrous oxides of iron and manganese have a pH-dependent surface charge which enables them to adsorb heavy metal ions (Jenne, 1968). Adsorption by organic matter is more complex. Organic matter can be regarded as having three components: humic acids (soluble in basic solutions), fulvic acids (soluble in both basic and acid solutions), and insoluble humins. There are no distinct divisions among these three, as all are part of a heterogeneous polymer system of molecules. Basic differences to distinguish between them are in elemental composition, acidity, molecular weight, and degree of polymerization. Reactive functional groups on these molecules include carboxyl, phenolic and alcoholic hydroxyls, methoxyl, carbonyl, and quinone groups. The negative charge of organic species is often due to the ionization of acidic carboxyl and hydroxyl groups. The charge can be neutralized by an interaction with a charged metal ion in solution. Attractive interactions could range from weak forces that make the metal ion easily replaceable (physical adsorption) to strong forces similar to chemical bonding (chemical or specific adsorption) (Jackson et al., 1978). Studies by Gibbs (1977) and others have established the importance of hydromorphic phases as significant scavengers of heavy metals in stream and lake sediments, and in numerous laboratory experiments. Experiments with 103 synthetic and natural sediment phases have identified a variety of adsorption behaviors for heavy metals. Krauskopf (1956) noted the importance of adsorption in controlling the distribution of a number of heavy metals in seawater, especially zinc, copper, and lead. The distributions of chromium and nickel were resolved to be controlled by organic reactions. Concentrations of heavy metals in freshwater sediments are significantly controlled by hydrous iron and manganese oxides (Jenne, 1968), organic molecules, and other hydromorphic phases such as clays, carbonates, and sulfides (Gibbs, 1977). Hydrous manganese oxides are able to adsorb metal ions from solution quite rapidly in laboratory experiments when pH and ionic strength are controlled variables. Posselt et al. (1968) found that exchange reactions are a principle mechanism in metal uptake. Group I metals undergo non- specific adsorption under electrostatic interactions. Manganese dioxides, which in colloidal hydrous form exhibit a negative surface charge within the pH range of 5 to 11, can have surface areas of 150 to 300 square meters per gram. The negative charge results from an increase in the ratio of (”17-bound to H+-bound ions, especially as pH increases. Heavy metals such as nickel, copper, and cobalt with smaller Crjystalline ionic radii than Group I metals (and therefore Steater hydrated ionic radii) appear to undergo specific adsorption in which the metal ions exchange with H+ ions and 104 form stronger, bidentate bonds with adjacent surface-bound hydroxyl ions. When zinc is adsorbed on manganese dioxide, it can replace Mn+2 in the lattice, on the basis of crystal field theory, or it can interchange with bound H+ sites, in which 2 moles of H+ are released per 1 mole of Zn+2 (Loganathan and Burau, 1973). However, although this suggests that crystal field stabilization energies control adsorption selectivity on manganese dioxides, the selectivity order of transition metals for manganese dioxides (Mn>Co>Cu>Zn>Ni) does not follow the Irving-Williams order (ZnCo>Fe>Mn) and CFT is therefore not the only factor involved in metal ion selectivity (Murray, 1975). Murray and Dillard (1979) investigated the oxidative properties of manganese dioxide. Co+2 can be readily oxidized to Co+3, for example, when adsorbed at the MnOZ- solution interface. However, Ni+2 cannot be oxidized unless it is present at very high concentrations. This suggests a mechanism for the geochemical separation of NI and Co in surface aqueous systems. Lead can be readily adsorbed by Mnoz and was hypothesized to be oxidized to form PbOz. However, McKenzie (1980) found no evidence for the oxidation of lead on nine synthetic manganese oxides, and attributed the: binding of lead to be caused by a special affinity of mahganese oxides for lead, such as was found for cobalt. Leaad appears to be more strongly adsorbed on many oxides théan copper and zinc. 105 Catts and Langmuir (1986) studied the applicability of the site—binding model of the electric double layer to the adsorption of copper, lead, and zinc by synthetic manganese dioxides. Adsorption was controlled by surface reactions with divalent metal cations up to a pH of 6. At greater pH, metal hydroxide-complex reactions with the surface were a better fit with the model. In general, predicted behaviors fit experimental adsorption data, but were limited by a lack of quantification in predicting the magnitude of adsorption after changes in solution conditions. They suggested that equilibrium binding constants would have to be determined for conditions found in natural systems in order to be modeled. Features to be considered include ionic strength, sorbent concentration, surface characteristics, and competing metal concentrations. Streams with actively precipitating manganese oxides were suggested to best represent the optimum natural system to test the model. Gadde and Laitinen (1974) found that there was appreciable adsorption of lead, cadmium, and zinc near the ZPC of hydrous manganese oxides. This was not expected since the surface charge at this pH is essentially neutral. This adsorption behavior was attributed to specific (non- eleectrostatic) adsorption reactions. Similar behaviors were 31:30 noted for copper, nickel, and cobalt adsorption on hY'c'irous manganese oxides in experimental studies, and also 501: lead and cadmium adsorption on hydrous iron oxides. 106 Metal adsorption can also be either enhanced or inhibited by the interaction of dissolved ligands and adsorbent surfaces. A ligand which is complexed with a aqueous metal ion may then be adsorbed at the surface and stay complexed with the metal to form a surface-ligand-metal adsorption complex. If a dissolved ligand is adsorbed onto a surface, it may also serve as a new adsorption site for aqueous metals on the surface or it may stabilize the surface binding site, thus inhibiting any adsorption of‘ metal i0ns. A ligand may form a nonadsorbing complex in solution and compete with a substrate for coordination with a metal ion. It has been suggested that heavy metal distributions in natural systems may be controlled by a humic compound coating on oxide surfaces rather than simple metal reactions with oxide surface binding sites (Davis and Leckie, 1978). The chemical modeling of heavy metal distributions is limited by the lack of knowledge of concentrations, identities of organic compounds, and stability constants for many metal-ligand and metal-surface interactions. Vuceta and Morgan (1978) examined chemical interaction and competition between metal ions and different components of natural wa‘tzers in an experiment with controlled parameters. They used oxic fresh water with four major cations (Ca, Mg, K, Na ), eight trace metals (Pb, Cu, Ni, Zn, Cd, Co, Hg, Mn, Fe ), eight inorganic ligands (C03, 304, c1, 3, Br, NH3, 90,, (”1) and a substrate with adsorption characteristics of 8102 107 (with calculated conversions to also represent Fe(OH)3 and MnOz). The investigation varied the following parameters: pH, types of adsorbing surfaces, surface area, and presence of organic ligands such as EDTA, amino acids, and citrate. This study stresses the importance of having a well- characterized environment in which to determine quantitative chemical adsorption behaviors of metals in natural aqueous systems. Vuceta and Morgan (1978) found that dissolved Cu+2 will be expected to be substantially removed from solution by adsorption if complexing agents are absent or in very low concentrations. If organic ligands present in a solution have little affinity for 9642, then the distribution of lead adsorption is dependent on the availability of substrate surface area. An increase in the substrate surface area and concentrations of organic ligands will cause an increase in the adsorption and complexation of Ni+2, Co+2, and Zn+2. It appears that the presence of both inorganic and organic complexing ligands may have a dramatic effect on heavy metal adsorption behaviors and on substrate surface Properties. Oxide surfaces can bind with either the metal or the ligand of a dissolved complex. By studying chloro- and 3L1 lfato- complexes of cadmium, Benjamin and Leckie (1982) fOund that if the metal end of a complex is adsorbed, a plot °EE percent metal adsorbed vs. pH is roughly parallel to that off a ligand-free system (but is shifted to a higher pH). If trle ligand end is adsorbed, an increased ligand 108 concentration increases the amount of metal adsorbed in one pH range, while metal adsorbed decreases in a higher pH range. The adsorption behavior of complexes is apparently independent of the type of adsorbent, since similar results were found for a variety of substrates. Solutions with high ionic strength would be expected to have a greater influence on the extent of adsorption than dilute freshwater. Using seawater, Balistrieri and Murray (1982) were able to estimate the binding energies of Cu, Pb, Cd, and Zn on goethite. The four metals showed no competition between each other for sites on the oxide. Although CO3’2, PO4’2, and 8102 had no effect on the adsorption of these metals, the concentration of Mg+2 and so"2 in solution did influence adsorption by competing with the metals for binding sites. The pH range of the "adsorption edge" for each metal ion is a function of substrate concentrations in a seawater system. In a study of the adsorption of heavy metals on humic acids, Kerndorf and Schnitzer (1980) found a different binding strength sequence of metals for each of three different pH readings, 5.8, 4.7, and 2.4 (humic acids are Insoluble at pH<6.5). At pH = 2.4, the order was 4.7, the order was "9 >Fe>Pb>Cu=A1>Ni>Cr=Zn=Cd=Co=Mn . At pH Hg =Fe=Pb=Cu=Al=Cr>Cd>Ni=Zn>Co=Mn. At pH 5.8, the order was "9=Fe=Pb=A1=Cr=Cu>Cd>Zn>Ni>Co>Mn. They could find no cC>rrelations between the affinities of humic acids and metal atlomic weights, atomic numbers, valencies, and crystal and 109 hydrated ionic radii. Apparently, metal ions compete with protons and with each other for the humic acid binding sites. Robinson (1982) investigated the partitioning of copper and zinc adsorbed to five substrate coatings in stream sediments to determine important residence sites for the metals and to study the influence of stream water pH and metal concentration on the partitioning. The results show that over 90% of the total copper and zinc determined by selective chemical extractions of the coatings reside in manganese and iron oxide phases. He also concluded that stream water pH and metal concentration variations were not important in controlling the partitioning of the metals among the coating phases. In the marine environment, the controlling mechanism of certain trace metal concentrations has been suggested to be adsorption by particles. Laboratory studies of adsorption often use natural particle phases or synthetic solids in well-defined solutions, such as seawater. The study of adsorption on synthetic solids has been primarily responsible for the development of surface adsorption theory. Balistrieri and Murray (1983) applied these concepts (which were developed using Cu, Cd, and Zn adsorption on well-defined goethite, FeOOH) to the interaction of Zn and adsorbents in natural sediments to determine apparent e, and Weber, W.J. 1968. Cation adsorption on colloidal hydrous manganese dioxide. Environmental Science and Technology, v. 2, p. 1087- 1093. Post, E.V., and Hite, J.B. 1964. - (map). USGS Mineral Resource Investment Map MF-278. Post, E.V., Lehmbeck, W.L., Dennen, W.H., and Nowlan, G.A. 1967. Stream_aediments.(map). USGS Mineral Resource Investment Map MF—301. Rendell, P.S., Bately, G.E., and Cameron, A.J. 1980. Adsorption as a control of metal concentrations in sediment extracts. Enviromental Science and Technology, v. 14, p. 314-318. Ringwood, A.E. 1955. The principles governing trace element distribution during magmatic crystallization. Geochimica et Cosmochimica Acta, v. 7, p. 189-202, 242- 254. Robinson, G.D., and Carpenter, R.W. 1979. Distinguishing significant from false copper and nickel anomalies in soil overlying the Gladesville Norite, Jasper Co., Georgia. Journal of Geochemical Exploration, v. 11, p. 157-173. Robinson, G.D. 1981. Adsorption of Cu, Zn, and Pb near sulfide deposits by hydrous Mn-Fe oxide coatings on stream alluvium. Chemical Geology, v. 33, p. 65-76. 129 Robinson, G.D. 1982. Trace metal adsorption potential of phases comprising black coatings on stream pebbles. Journal of Geochemical Exploration, v. 17, p. 205-219. Robinson, G.D. 1984/1985. Sequential chemical extractions and metal partitioning in hydrous Mn-Fe-oxide coatings: reagent choice and substrate composition affect results. Chemical Geology, v. 47, p. 97—112. Ruffner, J.A., and Bair, E.E. 1981. The Weather Almanac. 3rd Edition. Gale Research Co., Detroit, MI. Salomons, W., and Forstner, U. 1984. Meteie_in_rhe Hydrooyoiei,Springer-Verlag, New York. Schmidt, R.G. 1974. Preliminary study of rock alteration in the Catheart Mountain molybdenum-copper deposit, Maine. USGS Journal of Research, v. 2, No. 2, p. 189—194. Shuman, M.S., Haynie, C.L., and Smock, L.A. 1978. Modes of transport above and below waste discharge on the Haw River, N. Carolina. Environmental Science and Technology, v. 12, p. 1066-1069. Skougstad, D.E., Fishman. M.J., Friedman, L.C., Edmann, D-Eu and Duncan. 8.8- 1979. W Book 5. Chapter 1. in W inyeerioerione. USGS, Washington, D.C. Sokal, R.R., and Rohlf, F.J. 1969. Biomerry, Freeman and Co., San Francisco, CA. Stumm, W., and Morgan, J.J. 1981. Aooerio_gnemierry, 2nd Edition. Wiley Interscience, New York. Takacs, M.J., Gephart, C.J., and Long, D.T. 1983. The use of ternary diagrams for the interpretation of the physical and chemical sequestering of metals by stream sediment. (Abstr) Geological Society of America Abstracts with Programs, v. 15, No. 4. Tessier, A., Campbell, G.C., and Bison, M. 1979. Sequential extraction procedure for the speciation of particulate trace metals. Analytical Chemistry, v. 51, p. 844-850. Tessier, A., Campbell, G.C., and Bisson, M. 1980. Trace ’ metal speciation in the Yamaska and St. Francois Rivers (Quebec). Canadian Journal of Earth Science, v. 17, p. 90-105. 130 Tessier, A., Campbell, G.C., and Bisson, M. 1982. Particulate trace metal speciation in stream sediments and relationships with grain size: implications for geochemical exploration. Journal of Geochemical Exploration, v. 16, p. 77-104. Tessier, A., Rapin, F., and Carignan, R. 1985. Trace metals in oxic lake sediments: possible adsorption onto oxyhydroxides. Geochimica et Cosmochimica Acta, v. 49, p. 183-194. Thompson, W.B., and Borns, H.W. 1985. Snrfioiai_§eo1ogio_neo o£_neine (map). Maine Geological Survey, Augusta, ME. USGS. 1970. USGS, Washington, D.C. Van Valin, R., and Morse, J.W. 1982. An investigation of methods commonly used for the selective removal and characterization of trace metals in sediment. Marine Chemistry, v. 11, p. 535-564. Vuceta, J., and Morgan, J.J. 1978. Chemical modeling of trace metals in fresh waters: role of complexation and adsorption. Environmental Science and Technology, v. 12, p. 1302-1308. Wedepohl. LR. 1969. W Springer -Verlag, New York. Westall, J.C., Zachary, J.L., and Morel, F.M.M. 1976. MINEQL, a computer program for the calculation of chemical equilibrium composition of aqueous systems. Water Qual. Lab. Tech. Note No. 18, Dept. of Civil Eng., Mass. Inst. of Tech., Cambridge. Whitney, P.R. 1975. Relationship of manganese-iron oxides and associated heavy metals to grain size in stream sediments. Journal of Geochemical Exploration, v. 4, p. 251-263. Whitney, P.R. 1981. Heavy metals and manganese oxides in the Genesee watershed, New York: effect of geology and land use. Journal of Geochemical Exploration, v. 14, p. 95-117. Wolery, T.J. 1983. EQBNR - a computer program for geochemical aqueous speciation-solubility calculations: User's guide and documentation. UGRL-53414, Lawrence Livermore Natl. Lab., Livermore, CA. MICHIGAN STRTE UNIV. LIBRQRIES ‘N1|"“1“”lllWIW‘llWIIMIHIIWIHIHlI” H ”Ml 31293005330356