Metal dynamics in Lake Vanda (Wright Valley, Antarctica)

Chemical Geology, 76 (1989) 85-94 Elsevier Science Publishers B.V., Amsterdam - - Printed in The Netherlands

85

[6]

Metal dynamics in Lake Vanda (Wright Valley, Antarctica) WILLIAM J. GREEN 1, TIMOTHY G. FERDELMAN 1'2 and DONALD E. CANFIELD 3 ~School of Interdisciplinary Studies, Miami University, Oxford, OH 45056 (U.S.A.) 2College of Marine Studies, University of Delaware, Newark, DE 19716 (U.S.A.) 3Department of Geology and Geophysics, Yale University, New Haven, CT 06511 (U.S.A.) (Revised and accepted December 7, 1988)

Abstract Green, W.J., Ferdelman, T.G. and Canfield, D.E., 1989. Metal dynamics in Lake Vanda (Wright Valley, Antarctica). Chem. Geol., 76: 85-94. Data are reported for Mn, Fe, Co, Ni, Cu and Cd in the Onyx River, and for Mn, Co, Ni, Cu and Cd in Lake Vanda, a closed-basin Antarctic lake. Oxic water concentrations for Co, Ni, Cu and Cd were quite low and approximate pelagic ocean values. Scavenging of these metals by sinking particles is strongly indicated. Deep-lake profiles reveal a sharp peak in the concentrations of Mn, Fe and Co at the oxic-anoxic boundary at 60 m. Maxima for Ni, Cu and Cd occur higher in the water column, in the vicinity of a Mn submaximum, suggesting early release of these metals from sinking manganese oxide-coated particles. A rough steady-state model leads to the conclusion that there is a large downward flux of Mn into the deep lake and that this flux is sufficient to explain the annual loss of Co, Ni, Cu and Cd. A pronounced geochemical separation between Fe and Mn apparently occurs in this system - Fe being lost in near-shore environments and Mn being lost in deeper waters. Comparison of metal residence times in Lake Vanda with those in the oceans shows that in both systems Mn, Fe and Co are much more reactive than Ni, Cu and Cd. Energetically favorable inclusion of the more highly charged metals, Mn (IV), Fe (III) and Co ( III ), into oxide-based lattices is a plausible explanation.

1. I n t r o d u c t i o n

Geochemical research on closed-basin lakes has traditionally been directed at understanding how the major-element composition of lake brines evolves from the evaporation and transformation of surface and subsurface dilute inflows (Eugster and Jones, 1979). This approach has led to the articulation of general rules (Drever, 1982) governing the pathways taken by natural waters during evaporation. Considerably less attention, however, has focused on water-column processes involving the 0009-2541/89/$03.50

trace elements, and we have at present little understanding of how these elements are distributed, concentrated, scavenged and recycled in closed lake systems. Lake Vanda is, in a sense, an ideal system for examining metal behaviors. The lake is situated in Wright Valley (Antarctica), a glaciercarved valley bordered by the high rock walls of the Asgard and Olympus Ranges. It is 5.16 km long, 1.5 km wide and 68.8 m deep in the western depression (Nelson and Wilson, 1972 ), and is largely sequestered from atmospheric metal fluxes by a 4-m-thick permanent ice cover. Water is supplied by the Onyx River for roughly

© 1989 Elsevier Science Publishers B.V.

86

W.J. G R E E N E T AL.

a 6-week period from about mid-December to early February. The Onyx, which has its source at the Wright Lower Glacier 27 km to the east, has an annual discharge rate of ~2-109 1 (Chinn, 1981 ), and is likely to be, within a few percent, the single avenue of dissolved and suspended matter input. Vanda's distance from anthropogenic influence and its simple hydrologic regime recommend it as a system for examining material flows. Physically, the lake has been described as a two-layer system (Angino et al., 1965). The waters above 48 m are relatively fresh, while those below become rapidly more saline with depth. A convection current having a velocity of 1 cm s- 1was detected in the upper layer ( Ragotzkie and Likens, 1964 ) and is responsible for the homogeneity of the waters in this region. Representative features of the lake's vertical structure are shown in Fig. 1., where the existence of permanent well-defined oxic (above 55 m), suboxic (55-59 m) and anoxic (below 60 m) zones can be discerned. Details of the lake's geochemical evolution (Green and Canfield,

1984), nutrient chemistry (Canfield and Green, 1985), organic chemistry (Matsumoto et al., 1984) and biology (Vincent and Vincent, 1982 ) have been published recently. One of the distinguishing features of this lake is the existence of a natural diffusion cell below ~ 48 m. The CaC12 diffusion gradient for the bottom waters was first used by Wilson (1964) to date the most recent filling event for the lake at 1200 y. Canfield and Green (1985) have subsequently used diffusion modeling in their discussion of nutrient cycling in Vanda. The objectives of this study were: (1) to obtain vertical metal profiles having more detail than those reported earlier (Green et al., 1986) in order to better understand controls on metal concentrations in the lake; and (2) to estimate metal residence times so that residence time trends in this closed-basin system might be compared with trends in the ocean. For the sake of completeness, we include water-column Fe data from the 1980/1981 field work (Green et al., 1986). All other metal data are from the 1983/1984 field season.

pH 5

o

I

6

I

7

I

8

I

0"00

1 I

2 3 I I

~

! J~

5 I

O-@T log [(

20 _Ak-AHpS 25--lImpHq -

A

4 I

r

1015-

[CI-] mg/I

0 I

Ice

o

log

~k

30-

F

35-1/

~

I

(

50

,

q

~'40"1"65

6560 - - ~

p

i ~. .

.

• .

.

Zone

~ 1 ~ ~

.

I I I I I I ' '~ I ii 11 'l 'I il 0 8 1624324048-10001002003004005000 02* and H2S* (rag/I) Redox (mV) *From Torii et. al (1976)

i[

, 4

; ; ; ', 8 121620 Temp °C

I

I

2428

Fig. 1. Verticalprofilesof dissolved 02, pH, H2S, temperature, Eh (Pt electrode) and CI- in Lake Vanda. Sulfide data are from Torii et al. (1975).

87

METAL DYNAMICS IN LAKE VANDA

2. Methods

For this research, the Onyx River was sampled seven times during the period December 22, 1983, to January 10, 1984, and filtered and unfiltered river samples were analyzed for Mn, Fe, Co, Ni, Cu and Cd. Lake Vanda was sampled once at the deep site in late December 1983, and filtered and unfiltered samples were analyzed for Mn, Co, Ni, Cu and Cd. Lake samples were collected in an all-plastic acid-washed vertical sampler. These were either filtered immediately upon collection or transferred directly into 1-1 linear polyethylene bottles which had been soaked in 20% nitric acid for one week, in 1% Ultrex ® nitric acid for one day, and then rinsed with deionized glass-distilled water and with 1 1 of sample. Filters (0.45-/~m Millipore ® ) were soaked in HC1, rinsed with 1% Ultrex ® nitric acid (J.T. Baker Company) prior to transport to the field in separate acid-washed plastic containers, and cleaned further by filtering 1 l of water through them. Upon collection, waters were acidified to pH < 2 by addition of 2-3 ml of Ultrex nitric acid. River samples were treated in a similar manner. Metals were concentrated prior to analysis using a modification of the procedures of Danielson et al. (1978). The citrate buffer was cleaned by extracting three times with FreonTF ® (Aldrich Chemical Co.). A 1% w/v solution of ammonium l-pyrrolidine dithiocarbamate (APDC) and diethylammonium diethyldithiocarbamate (DDDC) was purifiedby double extraction into Freon-TF ®. This was prepared fresh daily. Freon-TF ® was cleaned by extracting three times with 3 ml of Ultrex ® HNO3 in 10 ml of ultrapure water. Lake waters were doubly extracted at pH ~ 5 into Freon-TF ® and then back-extracted into acid prior to analysis, resulting in a 20- to 40fold concentration depending on the sample. Mn was analyzed without extraction, using standard additions. Fe in river samples was determined after appropriate dilutions. Lake Fe concentrations were taken from the earlier work

of Green et al. (1986). All analyses were performed in triplicate using a Perkin-Elmer ® 3030 graphite furnace atomic absorption spectrometer with deuterium arc background correction. Procedural blanks were determined by extracting quartz-distilled water that had been acidified with 2 ml Ultrex ® nitric acid (see Table I), and these were subtracted from the sample values. Filtration under optimal laboratory conditions produced no additional signal. Some filtered samples, however, did show signs of contamination which we attribute to the handling of filters in the field under conditions of high winds and extreme cold. Filtered Mn values were in some cases higher than unfiltered values. In general, where comparisons are possible (Cu, Cd and Mn), the data presented here are in very good agreement with the less detailed water-column profiles presented by Green et al. (1986) for the 1980 season. The agreement lends credence both to the analytical results and to the view that metals in Lake Vanda are, like the nutrients (Canfield and Green, 1985), at or near steady state. 3. Results and discussion

Trace-metal data for the Onyx River (average of seven unfiltered samples) and for Lake Vanda are presented in Table I, and lake profiles are given in Figs. 2 and 3. Water-column profiles for unfiltered Ni, Cu and Cd exhibit a submaximum just beneath the ice sheet, with slowly decreasing concentrations down to ~ 40 m. Vestiges of the riverine plume, annual freezing-out of solutes which occurs during the Antarctic winter when ~ 30 cm of fresh ice is added to the bottom of the lake's ice cap (Chinn, 1981), and association of these metals with a sub-ice phytoplankton community may contribute to the high concentrations just below the ice-water interface. Particulate scavenging in the region between 20 and 40 m most likely leads to the concentration minima in this zone. Concentrations of all six metals exhibit peaks below 48 m, in the region of the lake characterized by stable thermal and chemical gradients. Be-

88

W.J. GREENET AL.

TABLE I

Metal concentrations in Lake Vanda and the Onyx River (water-column values for December 1983; except for Fe where values are for December 1980)

Onyx River .3 5m 15 25 35 45 48 51 54 56 58 60 62 64 66 % C.V. (3 injections) Detection limit Procedural blank

Fe (p,g l -~)

Mn (,ugl -x )

Co (ngl -~)

Cu (ngl -x)

Cd (ngl -~)

Ni (ngl -~)

U *l

U

F

U

F

U

F

U

F

U

F

24.8 0.87 0.68 1.44 2.28 3.43 7.76 88.4 585 513 422 3,670 2,740 3,210 3,400

15.7 0.70 0.82 2.19 2.28 3.94 7.13 90.2 610 530 428 3,785 2,670 3,030 3,030

628 <20 <20 < 20 < 20 <20 <20 <20 87 182 81 564 64 -~ 20 28

<20 <20 <20 < 20 <20 <20 <20 <20 92 71 32 < 20 < 20

517 2,606 1,434 1,269 392 807 565 1,190 3,987 12.010 11,330 11,400 13,230 1,774 4,594

111 2,433 112 <48 <48 <48 <48 <48 2,628 7,703 4,770 453 < 48 < 48 < 48

3.4 61 49 29 29 48 38 77 136 150 148 31 19 17 23

(3.4) *4 31 40 38 42 42 44 50 118 126 116 16 <1 <1

1,310 596 534 39 < 20 282 518 3,020 5,550 6,220 5,670 3,540 2,010 218 357

387 152 236 49 <20 122 489 n.a. 5,630 5,200 5,870 3,370 504

F .2

944 64 3.4 0.9 8.3 1.3 6.3 2.3 5.9 0.5 3.7 1.0 3.5 0.7 <0.1 <0.1 3.7 0.9 n.a. n.a. 12.4 4.1 1,040 960 n.a. n.a. 140

n.a.

10 0.1 < 0.1

10 0.05 < 0.05

12 20 86

8.0 48 126

11.0 1 7

79

7.0 20 132

- = suspected contamination; n.a. = no analysis. Fe at 57 m was 11.2/q~ 1-1 unfiltered and 4.1/~g 1-1 filtered. At 65 m, unfiltered Fe was 140 ~g 1-1; at 67 m values were 420/lg 1-1 unfiltered and 300/zg 1-~ filtered (see Green et al., 1986). *IU refers to unfiltered samples. *2F refers to samples filtered through 0.45-/~m filters upon collection. *3Onyx River average of seven samples. *4Filtered Cd assumed equal to unfiltered Cd.

low this depth there is a sharp increase in ionic strength, temperature and density. A photosynthetic maximum lies above the anoxic zone at 58 m (Vincent and Vincent, 1982), while a band of nitrifying bacteria has been found at 54 m (Vincent et al., 1981). The metal profiles fall into two discrete patterns. Fe, Mn and Co exhibit dissolved maxima across the redoxcline, suggesting reduction and dissolution of their respective oxide phases. We note that in the suboxic waters between 57 and 60 m, where the pH is ~ 6, Fe and Co concentrations increase by nearly two orders of magnitude and then decrease rapidly in the sulfidebearing waters below 60 m. Mn concentrations, by contrast, remain high throughout the anoxic zone. Pronounced similarities in the Mn and Co profiles between 50 and 60 m suggest that Co is

strongly associated with Mn geochemistry in this region. A particularly significant feature of the Mn profile is the submaximum at 54 m. We attribute this to the onset of reductive dissolution in suboxic low-pH waters, possibly according to the reaction: Mn02(,) -{-2H+~.-~-½02(g)+ M n 2+ +H20(l ) (I) The slight decrease in dissolved Mn below this depth may be due to the roughly coincident submaximum in dissolved oxygen (Fig. 1 ). This would cause reaction (I) to shift to the left. The Ni, Cd and Cu profiles are distinguished by dissolved maxima in the vicinity of this Mn submaximum. This may be the result of metal release from dissolving Mn-oxide particles and coatings in suboxic waters, and subsequent for-

89

M E T A L D Y N A M I C S IN L A K E V A N D A

0 F~ < ~

ICE r j

~,

H

^~ ~

~ 4 "

• ~ I C E <~ ~ ~ ~

lO

" 7 p ,¢

TM

"

P

Cu

L ~J

co

30)

60-

i

70 0

i

t

=

6000

ng1-1

0

I

I

I

12000

ng1-1

I

150 ng1-1

Fig. 2.Filtered and unfiltered concentrations of Cd, Ni and Cu in the Lake Vanda water column. "particulate" metal.

. . . .

"

tl

S h a d e d area

P • mvh= -- •

represents

. l

Co

Fe

Mn

10

I

0

20 E 3oe,. Q.

40) ) ) ) Y

50-

60-

70

I

0

I

pgr I

I

3000

I

0

I

#gl 1

I

1200

0

Fig. 3. Filtered and unfiltered concentrations of Mn, Fe and Co in the Lake Vanda water column.

I

I

ng1-1

u

600

90

mation of soluble complexes in the waters of the chemocline. Particles descending into this zone experience the formation of an abrupt pycnocline, bands of relatively high biological productivity, an increased availability of potential inorganic and organic ligands, and decreasing pH- and pe-values. Matsumoto et al. (1984) have observed peak organic concentrations at 55 m corresponding to the Ni, Cd and Cu dissolved maxima. Solubilization processes may also be enhanced in this region since the pycnocline effectively lengthens the residence time of sinking particles. The 60-m depth represents the transition from suboxic to anoxic conditions. The depletion of Fe, Co, Ni, Cu and Cd below this depth is due to the formation of insoluble sulfide phases in waters where total dissolved sulfides of 1.2.10 -3 M have been measured at 65 m by Torii et al. ( 1975 ). (See our calculations for Cu and Fe in Green et al., 1986.) Profiles similar to these have been observed in sulfidic marine environments, including the Black Sea (Brewer, 1975) and, more recently, Framvaren Fjord, Norway (Jacobs et al., 1985), where it was noted that even those metals which readily form bisulfide complexes (Cd(II) and Cu(I)) are removed from solution. A striking contrast exists, however, between the behavior of Ni in Framvaren Fjord and its behavior in Vanda. In the fjord, Jacobs et al. (1985) noted that Ni concentrations remained unaffected by the presence of high sulfide concentrations ( ~ 8 mM), and that vertical profiles showed no tendency for Ni to decrease with depth. In Vanda, on the other hand, there is a rapid decrease in Ni below 58 m. More extensive organic complexation of Ni in the higher-pH waters of the fjord may be responsible for its apparent supersaturation with respect to known sulfide phases. In lowerpH Vanda waters, protonation of potential organic chelates would lessen the tendency of these molecules to complex with Ni. We note in passing that we have observed Ni profiles in Lake Fryxell, a permanently anoxic Antarctic lake, which resemble those of Framvaren Fjord.

W.J. GREEN ET AL.

Anoxic Fryxell waters are in the pH range 7.37.5 (Green et al., 1989). Some idea of the rate of metal removal from this hydrologically closed system can be obtained from residence times, as computed here from unfiltered river inputs and unfiltered lake (down to 48 m) concentrations and bathymetry. The residence time t is defined according to:

t-~cY/q)ici

(2)

where c is the metal concentration in a designated compartment of the lake; Vis the volume of the compartment; 0i is the rate of stream inflow; and ci is the metal concentration in the inflow. Loadings were estimated from average metal concentrations and from the 10-yr. average Onyx River fluxes reported by Chinn (1981). The quantity of metal in the lake between 4 and 48 m was determined by multiplying the volume of successive conic sections by the average of the unfiltered metal concentrations at these depths and then summing over the entire lake. For all metals, upward diffusional fluxes from the brine into the 4-48-m compartment were insignificant ( < 2% ) as compared with the riverine flux. Residence times, computed in the above manner, are presented in Table II and are in the order: Cd >> Cu >> Ni > Mn > Co > Fe The position of an element in this series is best construed as an indicator of its relative reactivity in the lake. The order shows that the elements Mn, Co and Fe are extremely reactive in oxic waters and are quickly removed to the sediments or to the deep brine. The large difference between filtered and unfiltered Co concentrations in the Onyx River indicates that this element is bound up in riverine particulate phases. We suspect that most of the Co is lost to the sediments very soon after the Onyx emp-

91

METALDYNAMICSIN LAKEVANDA TABLE II Residence time of metals in the 3-48-m zone of Lake Vanda Metal

Mn Fe Co Ni Cu Cd

Unfiltered load (gyr. -1)

Filtered load (gyr, -1)

Upper lake content (g)

Residence times* (yr.) U

5.0" 104 1.9" 106 1.3" 103 2.6.103 1.0" 103 5.4

3.2 • 104 0.13-106 < 0.04.103 0.77" 103 0.21.103 5.4

3.9-105 8.3.105 3.6" 103 6.1.104 2.4.105 8.0.103

7.8 (9.9) 0.44 (1.4) < 2.8 23 240 (174) 1,500 (82)

12 6.4 < 90 79 1,100 1,500

*U refers to residence time calculated using unfiltered load; Fis residence time using filtered load. Residence times in parentheses are from the 1980 field season and are taken from Green et al. (1986).

ties into Lake Vanda and that little is actually transported to the body of the lake. Cd, on the other hand, has a much longer residence time and may be undergoing recycling within the lake's biologically community. In general, those elements which are incorporated into oxidebased lattices as 2 + ions have longer residence times than elements with higher solid phase charges (i.e. Mn(IV), Co(III), Fe(III)). The residence times for Mn, Fe and Cu are in good agreement with those reported earlier by Green et al. (1986). Values for the earlier study are given in Table II in parentheses. The Cd residence time, however, is considerably greater than the earlier value and we can do no more at this point than bracket t for this element. Residence times for both field seasons are plotted in Fig. 4 against Halbach's (1986) recent estimates of oceanic residence times. Although there is some scatter, it is clear that for both of these hydrologically closed systems, the transition metals Mn, Fe and Co have quite short residence times, while Ni, Cu and Cd tend to remain in the water column for a longer period. In the absence of data on the composition and vertical fluxes of particles, we can only speculate on the nature of trace-metal transport processes at this time. However, the vertical profiles do provide a clue. Table III is an estimate of annual metal loss from the anoxic zone, based on the downward

4

J

g, 3

Log tvanda Log toceans

--I

®Cu

2

t Cd

®

Ni

1 Co 0 -10

P

Mn

~Fe 1

2

3

4

5

6

Log tocean$ Fig. 4. Residence times of metals in Lake Vanda vs. oceanic residence times.

diffusional flux. If these are steady-state profiles, then the loss must be balanced by an annual influx of metal from the upper water column. What is striking in Table III is the large calculated Mn flux. We have since (1986/1987 field season) verified both the Mn submaxim u m near 54 m and the decrease in Mn below 60 m. The gradient is within ___10% of that reported here. We estimate that 1.2-104 g y r . - ' Mn must be raining down into the anoxic zone to balance the diffusional loss of this metal. This is roughly 37% of the dissolved annual river load. By contrast, only 1.6% of the dissolved Fe load is required to compensate for Fe diffusional losses. This calculation suggests that there is a geo-

w.J. GREENET AL.

92 T A B L E III Annual fluxes through 60-m contour and comparison with river fluxes Metal

Mn 2+ Fe 2+ Co 2+ Ni 2+ Cu 2+ Cd 2+

D (18°C) (10Scm2s -1)

Depthsfor obtaining gradient

Gradient3c/Az Area of 60-m Annual flux (10s/~gcm -4)

contour (10-1Ocm 2

Filtered % river load to sediments river load lost through (gyr. -1) (gyr. -1) 60-mcontour

Required wt.% of minor element in MnO2 phase

5.75 5.80 5.72 5.81 5.88 6.03

60-62 60-67 60-62 58-62 56-60 58-62

558 94 0.3 1.3 1.8 0.25

1.2 1.2 1.2 1.2 1.2 1.2

1.2 "104 0.21-104 6.0 29 40 5.7

-

3.2.104 37.5 13 .104 1.6 <40 < 15 700 4.1 210 19 5.4 < 100

0.03 0.15 0.21 0.03

Values in last column are much lower than metal concentrations in seamount precipitate; 60-m contour is 24.4% of water surface at 3 m.

chemical separation occurring between Mn and Fe, such that a significant fraction of the incoming Mn (37%) is being lost in the deep lake environment, while most of the Fe is being lost near shore. This is shown more clearly on the schematic bathymetric map in Fig. 5. Given the large Mn flux to deep lake waters, it seems reasonable that manganese oxides and hydrous oxides are acting as the principal scavengers and transport agents for the metals Co, Ni, Cu and Cd. This point is reinforced by the similar vertical profiles exhibited by Co and Mn between 50 and 60 m. We attribute the Co submaximum at 56 m and the maximum at 60 m to the dissolution of manganese oxide carrier phases, and to subsequent release of lattice-

bound Co (III). The shallower maxima for Ni, Cu and Cd suggest that these elements are transported in less strongly bound forms and are being released at the onset of particulate Mn dissolution near 55 m. In the last column of Table III we note that the minor metals would have to constitute only a small weight fraction of the sinking Mn particles (assumed to be MnO2 for this calculation) in order to have their downward fluxes fully accounted for. The percentages of Co, Ni, Cu and Cd estimated here are much lower than the corresponding percentages of these metals in marine Mn minerals (Halbach, 1986). A test of these conjectures, however, will require information to be obtained from recent deploy-

=

Fe Removal

Fig. 5. Schematic bathymetric map showing fluxes of Mn and Fe through 60-m contour.

Onyx River

METALDYNAMICSIN LAKEVANDA

ment of particle traps at various locations in the lake.

4. Conclusions The Onyx River transports ~2000 kg Fe, ~ 50 kg Mn, ~ 1.6 kg Ni, 1 kg each of Co and Cu, and ~ 5 g Cd to Lake Vanda on an annual basis. Only a small fraction of the Fe (7%) and Co ( < 3% ) is "dissolved", but for Mn, Ni, Cu and, presumably, Cd, the fractions of dissolved metal are much higher. In spite of the fact that Vanda is a hydrologically closed system, metal concentrations in the upper (4-48 m) oxic water column are quite low. At 48 m, for example, Ni, Cu and Cd levels (8.3, 8.9 and 0.3 nM, respectively) are similar to concentrations reported in pelagic ocean water (Bruland, 1980). These low values suggest that the water column is being cleansed by effective metal scavenging processes. Vertical profiles in the deep lake exhibit two clear patterns. Mn, Fe and Co show sharp peaks at 60 m, which suggests that sinking oxide and hydrous oxide particles are undergoing reduction and dissolution in the low-pe,low-pH waters at this depth. Similarities in the shapes of the Mn and Co profiles indicate that manganese oxides may be acting as the principal carrier phases for Co - bound perhaps as crystal-field-stabilized Co (III) in the oxide lattice (Burns, 1976). For Ni, Cu and Cd, maxima occur higher in the water column; transport and release of these metals by manganese oxide phases are strongly indicated. Analysis of downward diffusional fluxes and the assumption of steady-state lead to the conclusion that there is a large Mn flux through the 60-m contour of the lake, amounting to ~ 37% of the dissolved annual river flux. This calculated rain of Mn through the water column is considered sufficient to account for the annual loss of Co, Ni, Cu and Cd from the deep lake. By contrast with Mn, Fe appears to be lost largely in near-shore environments. Residence times for Mn, Fe and Cu were in

93

good agreement with our earlier estimates, but the Cd residence time was much longer in this study. A plot of metal residence times in Lake Vanda, against recent estimates of oceanographic residence times, reveals that in both systems the more highly charged metals, Mn, Fe and Co are removed more rapidly than Ni, Cu and Cd. Future studies of Lake Vanda include the deployment of particle traps as a way of testing the conjectures advanced in this paper.

Acknowledgements We are grateful to the National Science Foundations Office of Polar Programs for financial support. Mike Angle made a significant contribution to this work in Wright Valley, as did the pilots and crew of VXE-6. We thank Professor Bud Williamson of Miami University for equipment support and for his general encouragement of this research. Conversations with Gunter Faure and Keith Chave have added materially to our understanding of the Dry Valley Lakes. We thank Betty Marak for her careful work in preparing the manuscript.

References Angino, E.E., Armitage, K.B. and Tash, J.C., 1965. A chemical and limnological study of Lake Vanda, Victoria Land, Antarctica. Univ. Kans. Sci. Bull., 45: 10971118. Brewer, P.G., 1975. Minor elements in seawater. In: J.P. Riley and G. Skirrow (Editors), Chemical Oceanography, Vol. 1. Academic Press, London, pp. 415-496. Bruland, K.G., 1980. Oceanographic distributions of cadmium, zinc, nickel, and copper in the North Pacific. Earth Planet. Sci. Lett., 47: 176-198. Burns, R.G., 1976. The uptake of cobalt into ferromanganese nodules, soils and synthetic manganese (IV) oxides. Goechim. Cosmochim. Acta, 40: 95-102. Canfield, D.E. and Green, W.J., 1985. The cycling of nutrients in a closed-basin Antarctic lake: Lake Vanda. Biogeochemistry, 1: 233-256. Chinn, T.J.H., 1981. Hydrology and climate in the Ross Sea area. J.R. Soc. N.Z., 11: 373-386. Danielson, L.G., Magnusson, B. and Westerlund, S., 1978. An improved metal extraction procedure for the deter-

94 mination of trace metals in seawater by atomic absorption spectrometry with electrothermal atomization. Anal. Chim. Acta, 98: 47-57. Drever, J.I., 1982. The Geochemistry of Natural Waters. Prentice-Hall, Englewood Cliffs, N.J., 388 pp. (see especially pp. 200-228). Eugster, H.P. and Jones, B.F., 1979. Behavior of major solutes during closed-basin brine evolution. Am. J. Sci., 279: 609-631. Green, W.J. and Canfield, D.E., 1984. Geochemistry of the Onyx River (Wright Valley, Antarctica) and its role in the chemical evolution of Lake Vanda. Geochim. Cosmochim. Acta, 48: 2457-2467. Green, W.J., Canfield, D.E., Lee, G.F. and Jones, R.A., 1986. Mn, Fe, Cu, Cd distributions and residence times in closed-basin Lake Vanda (Wright Valley, Antarctica). Hydrobiologia, 137: 237-248. Green, W.J., Gardner, T.J., Ferdelman, T.G., Angle, M.P., Varner, L.C. and Nixon, P., 1989. Geochemical processes in the Lake Fryxell Basin (Victoria Land, Antarctica). Hydrobioiogia, 172: 129-148. Halbach, P., 1986. Processes controlling the heavy metal distribution in Pacific ferromanganese nodules and crusts. Geol. Rundsch., 75: 235-247. Jacobs, L., Emerson, S. and Skei, J., 1985. Partitioning and transport of metals across the O2/H2S interface in a permanently anoxic basin: Framvaren Fjord, Norway. Geochim. Cosmochim. Acta, 49: 1433-1444.

w.J. GREENET AL. Li, Y.-H., 1981. Ultimate removal mechanisms of elements from the ocean. Geochim. Cosmochim. Acta, 45: 16591664. Matsumoto, G., Torii, T. and Hanya, T., 1984. Vertical distribution of organic constituents in an Antarctic lake: Lake Vanda. Hydrobiologia, 111: 119-126. Nelson, C.S. and Wilson, A.T., 1972. Bathymetry and bottom sediments of Lake Vanda, Antarctica. Antarct. J. U.S., July-Aug., pp. 97-99. Ragotzkie, R.A. and Likens, G.E., 1964. Heat balance of two Antarctic lakes. Limnol. Oceanogr., 9: 412-425. Torii, T., Yamagata, N., Nakaya, S., Murata, S., Hashimoto, T., Matsubaya, O. and Sakai, H., 1975. Geochemical aspects of the McMurdo saline lakes with special emphasis on the distribution of nutrient matters. In: T. Torii (Editor), Geochemical and Geophysical Studies in the Dry Valleys, Victoria Land, Antarctica. Mem. Natl. Inst. Polar. Res. (Jpn.), Spec. Iss. No. 4, pp. 5-19. Vincent, W.F. and Vincent, C.L., 1982. Factors controlling phytoplankton production in Lake Vanda. Can. J. Fish Aquat. Sci., 39: 1602-1609. Vincent, W.F., Downes, M.T. and Vincent, C.L., 1981. Nitrous oxide cycling in Lake Vanda, Antarctica. Nature (London), 292: 618-620. Wilson, A.T., 1964. Evidence from chemical diffusion of a climate change in the McMurdo Dry Valleys 1200 years ago. Nature (London), 201: 176-177.