Rare earth elements in river waters

Earth and Planetary Science Letters, 89 (1988) 35-47

35

Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

[6]

Rare earth elements in river waters Steven J. G o l d s t e i n * a n d Stein B. J a c o b s e n Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA 02138 (U.S.A.) Received April 1, 1987; revised version received March 21, 1988 We measured rare earth element (REE) concentrations in river waters to characterize the suspended and dissolved river flux of the REE to the oceans. The REE pattern of river suspended materials is sensitive to drainage basin geology. A positive correlation is observed between L a / Y b ratios and Nd model ages for the rivers studied. Major rivers have light REE enriched patterns relative to the North American Shale Composite (NASC), with ( L a / Y b ) N = 1.6-2.7. River water dissolved material ( < 0.2 #m) is heavy REE enriched relative to suspended material, and the most pronounced negative Ce anomalies occur in rivers of high pH. Light REE concentrations vary by approximately 3 orders of magnitude and are inversely related to p H and major cation concentrations. From these data, we estimate that typical major river runoff has heavy REE depleted suspended material with ( L a / Y b ) N = 1.9. We conclude that the terrigenous input to the oceans from major rivers is heavy REE depleted relative to shales. From the available data, average river water dissolved material appears to be heavy REE enriched with ( L a / Y b ) ~ --- 0.4. Estuarine removal processes lower the dissolved REE river flux by approximately 60% and result in a flux that is more heavy REE enriched with ( L a / Y b ) r ~ = 0.2. Calculated oceanic residence times with respect to river input range from 2300 to 21,000 years, are shortest for Ce, and greatest for the heavy REE and La. Such long residence times may suggest the presence of additional sources of REE in seawater.

1. Introduction Rare earth d e m e n t (REE) abundance patterns of river water dissolved and suspended material should provide insight into both the seawater cycle of R E E and the abundances of these elements in the continental crust. Neutron activation data on the suspended load of 5 major rivers [1] suggest a flat or slightly light R E E enriched pattern relative to shales. Dissolved load data are available for only a few rivers [1-3] and suggest a flat or slightly heavy R E E enriched pattern relative to shales. On the basis of these sparse data, m a n y authors have assumed that the river input to the oceans is characterized by a flat REE pattern relative to shales [4-7]. Clearly, more data are necessary to characterize the input to the oceans of REE in the dissolved and suspended load of rivers. In an attempt to

* Present address: Los Alamos National Lab., Los Alamos, N M 87545, U.S.A. 0012-821X/88/$03.50

© 1988 Elsevier Science Publishers B.V.

rectify this situation we have measured the R E E abundances of dissolved ( < 0.2 ~m) and suspended material from 9 rivers. F r o m these data we estimate the typical R E E pattern of the dissolved and suspended load of average river water. The implications of these data for the behavior of R E E during weathering and river transport, the marine geochemical cycle of the REE, and the R E E systematics of the continental surface are briefly discussed.

2. Samples and methods Samples of Amazon, Indus, Mississippi, Murray-Darling, and Ohio rivers were selected because they are all rivers with large drainage areas or discharges. In addition, a few samples of smaller rivers were selected to establish the R E E pattern of rivers with more distinct drainage basin lithology and water chemistry. These samples include Lake Isua in West Greenland and the Great Whale River in northwest Quebec; two dilute waters of low p H which drain Archean metasediments and

36 gneisses, and the Pampanga River in the Philippines and the Shinano River in Japan, two rivers which drain young island arc terranes. The locations and dates of the samples have been given elsewhere [8] with the following exceptions: The Indus River sample was obtained at Partab Bridge on 01-16-86. The Amazon River particulate sample was a river-bottom sample obtained on 05-22-85 at station MTCS 8 located in the river plume near the mouth of the river. The Amazon River dissolved load sample ( # $207; [9]) was obtained on 06-13-76 near Apixuna. Samples of Great Whale, Mississippi, Murray, Ohio, Pampanga, and Shinano rivers were all filtered in the field through 0.2 # m Millipore filters. The Isua and Indus samples were filtered through 0.2 /~m filters 10 and 90 days, respectively, after sample collection. The Amazon dissolved load sample was filtered through a 0.45 # m Nucleopore filter in the field. Dissolved load samples were acidified with ultrapure HC1 to prevent adsorption during storage. The dissolved R E E in these 100-400 g samples were spiked with a mixed REE isotopic tracer and concentrated by coprecipitation with ferric hydroxide. The particulate samples of 1 to 60 mg were ashed for approximately 1 hour at 850 ° C in a clean platinum crucible and spiked with a mixed R E E isotopic tracer. The mixed REE isotopic tracer was calibrated to better than 0.1%. This was accomplished with standard solutions made from ultrapure R E E metals. Particulate samples of the North American Shale Composite and the Amazon River were dissolved in H F in teflon bombs, whereas the other samples were dissolved in HF, HC1, H N O 3, HC104 in an open teflon beaker. No residues were observed upon reheating the samples in HC1. The REE were separated from Fe, Ba, and the major cations by ion chromatography using a gradient elution scheme [10]. The REE were loaded onto a triple filament assembly and analyzed by sequential ionization [11]. La, Ce, Nd, Sin, Eu and Dy concentrations are determined to better than 2%, while Gd, Er, Yb and Lu are determined to better than 5%. Chemistry blanks are ~<100 pg for N d and are ~<10% of the smallest sample (Indus River dissolved load). For all other samples blanks are ~< 2%. The REE concentrations measured in this laboratory in the rock standard BCR-1 and in the

North American Shale Composite (NASC) [12] are given in Table 1 and were measured with an uncertainty ~< 2%. Our measured R E E concentrations for BCR-1 are generally within 0.5% of the consensus reference values for this standard [13]. The best REE concentrations reported previously for NASC are those of Gromet et al. [12]. They measured Ce, Nd, Sm, Eu, Gd, Dy, Er, and Yb by isotope dilution mass spectrometry (IDMS) of a sample dissolved by H F b o m b dissolution. Their La and Lu concentrations were measured by instrumental neutron activation analysis. Except for the Ce concentration of Gromet et al. [12], which is within 2% of our value, their values are 9% to 19% higher than the values we report for Nd to Lu in Table 1. Their La concentration is 9% lower than the values reported in Table 1 and thus, the L a / Y b ratio measured for N A S C in our laboratory is 28% greater than the value reported in [12]. We also obtained ENd(0 ) = - - 1 4 . 7 _+ 0.3, a47Sm/144Nd=0.1167, and T ~ = 2 . 0 0 G a for NASC, in agreement with the previous measurement by McCulloch and Wasserburg [14]. The most likely cause of the difference in the REE concentrations for N A S C as measured in our laboratory and in [12] is heterogeneity of the NASC powder. As reported in [12], the reaction of NASC powder and H F in an open beaker is not sufficient to dissolve the heavy REE completely. Because our N A S C powder was reacted with H F in a teflon lined stainless steel b o m b at 180°C, and no residues were apparent after dissolution, we conclude that incomplete dissolution of N A S C is not the cause of the differences in R E E concentration reported by us and [12]. G r o m e t et al. [12] reported a large range (up to 15%) in R E E concentrations for different aliquots of N A S C powder which have been subjected to the same method of dissolution and measurement. For example, the N d / Y b ratio reported in [12] for I D M S ranges from 9.6 to 10.7. This range is larger than the analytical precision of I D M S and suggests that the NASC powder is heterogeneous with respect to R E E concentration [12]. Our measured N d / Y b ratio is 11.3, which is not significantly different from the range of values reported by [12]. Consequently, the differences between our values and those of [12] probably reflect to a large extent the heterogeneity of the N A S C powder rather than any systematic analytical difficulties.

37 2.5

3. Results

o -• ~ --

o Amazon • Great Whale ~ Indus • - - A Lake Isua D -E~ MlssllgIppl

A--A

o

R E E concentrations in the dissolved and suspended load of the rivers studied are presented in Table 1. R E E concentrations measured for N A S C in this laboratory are used for data normalization and are also presented in Table 1.

co

• I P - - - • ~

•--

• Murroy

9 - -

V

• --

D

•

o

Ohio • Pampanga o Shl o

D~

"; ~

3.1. Suspended loads The NASC-normalized R E E patterns of river water suspended material are plotted in Fig. 1. The two smaller rivers that drain Archean terrains, Lake Isua and the Great Whale River, have patterns that are strongly enriched in the light REE. The major rivers in this study have patterns that are moderately enriched in the light REE. The two rivers which drain young island arcs, the Shinano and Pampanga, have flat and heavy R E E enriched patterns, respectively. M a n y of these rivers have a small positive Eu anomaly. Ce behaves similarly to the other light REE.

2.0

.

1.0

.

.

.

.

0.5 0.0

i

~

Lo Ce

i

i

Nd

i

i

i

i

i

i

Srn Eu Gd

Dy

i

L

i

t

Er

i

Yb Lu

Fig. 1. N o r t h A m e r i c a n Shale C o m p o s i t e ( N A S C ) n o r m a l i z e d R E E a b u n d a n c e s in river s u s p e n d e d loads. Rivers d r a i n i n g island arcs have flat to heavy R E E enriched p a t t e r n s , a n d rivers d r a i n i n g A r c h e a n terrains have strongly light R E E enriched patterns. M a j o r rivers have l i g h t - R E E - e n r i c h e d p a t t e r n s with ( L a / Y b ) N = 1.4-2.7.

The NASC-normalized L a / Y b ratio: (La/Yb)N

-

(La/Yb)sample//(La/Yb)NAS

C

(1)

varies from nearly 0.22 for the Pampanga River to

TABLE 1 R a r e e a r t h element c o n c e n t r a t i o n s in river waters Sample

La

Ce

Nd

Sm

Eu

Gd

Dy

Er

Yb

Lu

Weight a

S u s p e n d e d loads b

Amazon G r e a t Whale Indus Isua-F Mississippi Murray Ohio Pampanga Shinano

35.0 52.1 19.4 72.8 43.7 37.5 41.2 7.72 28.6

72.9 103 41.4 143 92.5 71.4 83.9 18.2 62.7

32.6 38.9 18.6 52.0 39.8 35.2 36.9 12.6 27.0

5.93 5.79 3.70 8.01 7.52 6.98 6.91 3.55 5.75

1.10 1.11 0.89 1.08 1.52 1.56 1.39 1.14 1.24

74.0 227 2.91 609 19.7 6.31 4.28 37.4

212 337 2.41 1220 9.67 10.4 9.48 83.4

127 167 3.20 452 19.9 10.8 8.60 49.7

34.5 23.8 0.71 72.4 4.50 2.54 2.49 11.0

7.90 3.82 0.22 10.8 1.11 0.66 0.82 2.62

34.0 25.0

66.7 53.6

30.1 30.1

5.80 6.55

1.16 1.92

4.18 -

3.02 5.54 5.93 5.73 5.13 4.89 5.35

2.64 2.94 2.48 3.70 5.05 4.64 3.95 4.68 4.65

1.23 1.46 1.24 1.49 2.44 2.10 1.89 2.85 2.52

1.02 1.32 1.07 1.35 2.09 1.84 1.54 2.74 2.33

0.15 0.20 0.16 0.18 0.29 0.28 0.41 0.36

31.4 11.1 1.25 36.3 7.56 5.63 3.89 12.1

16.6 5.76 0.95 17.5 6.53 4.53 2.92 7.37

15.3 5.75 0.94 14.4 6.06 3.61 2.72 7.09

0.94 0.17 2.11 0.58 1.60

4.67 6.37

2.73 3.71

2.67 3.39

0.41 -

57.95 0.87 1.85 20.28 17.80 17.27 38.81 18.55 20.34

Dissolved loads c

Amazon G r e a t Whale Indus Isua-F Mississippi Ohio Pampanga Shinano

-

50.3 -

-

-

-

Standards d

NASC BCR-1 a b ¢ d

5.12 6.82

58.07 113.09

W e i g h t of dissolved sample in mg. S u s p e n d e d load R E E c o n c e n t r a t i o n s in p a r t s per million (ppm). Dissolved l o a d R E E c o n c e n t r a t i o n s in p a r t s p e r trillion (ppt). R E E c o n c e n t r a t i o n s ( p p m ) in the N o r t h A m e r i c a n Shale C o m p o s i t e (NASC) a n d the U S G S s t a n d a r d BCR-1 as m e a s u r e d at Harvard.

38 TABLE 2 Rare earth element fractionations River

Discharge a

Amazon Ohio Mississippi Indus Great Whale Murray Shinano Pampanga Isua-F

dl sl KLa/Y b

Dissolved load

6930 162 153 94 38 32 22 14 0.02

(La/Yb)~

Ce*

pH

(Ca + Na) (ppm)

0.38 0.14 0.25 0.24 3.1 0.41 0.12 3.3

1.1 0.64 0.24 0.39 0.80 0.97 0.79 1.1

6.62 7.85 8.05 7.8 6.80 8.02 7.11 7.86 6.0

6.4 56.1 81.2 66.0 2.5 63.4 14.1 40.4 2.0

0.14 0.07 0.16 0.17 1.0 0.43 0.56 0.73

Suspended load (La/Yb) N

TI~MNdb

2.7 2.1 1.6 1.4 3.1 1.6 0.96 0.22 4.6

1.54 1.75 1.59 1.62 2.91 1.24 1.19 0.55 3.4

a Runoff (km3/yr) data from Korzun [17]. b Sm-Nd depleted mantle model age (Ga) [15,16].

4.5 for Lake Isua (Table 2). The suspended loads of the major rivers have values of ( L a / Y b ) N that vary within the small range of 1.59 to 2.70. Thus, it is apparent that river water suspended loads generally do not have flat REE patterns relative to shales. In Fig. 2, L a / Y b is plotted versus Tiff, the Sm-Nd depleted mantle model age of these materials (Table 2; [15]). The Sm-Nd depleted mantle model age yields an estimate of the mean age of the crustal source areas of the suspended loads. This figure indicates that suspended materials derived from rocks of young model age tend to have lower L a / Y b than suspended materials derived from older rocks. We have previously re-

60

. . . . . . .

5O

4o

>.-

30

,

.

O Amazon • Great Whole ZS Indus • Lake Isua [ ] Mississippi • Murray '~ Ohio • Pompanga 0 Shinono

.

.

.

.

.

.

.

.

,

.

.

.

.

.

.

.

.

.

r

. . . . . . . . .

•

• C)

ported a relationship between the S m / N d ratio and ToY~ for river water suspended material from a much larger set of rivers [15]; in these rivers the S m / N d ratio of the suspended load decreases as Tff~ increases. Consequently, both of these relationships suggest that the R E E pattern of river water suspended material changes systematically with the age of the rocks in the drainage area. Whereas the REE patterns for the smaller rivers in this study exhibit a large degree of variation, the major rivers appear to average out small scale variations, as they have relatively uniform R E E patterns. The relatively large range in the absolute abundances of the REE in these rivers is probably attributable to variations in the mineralogy of the suspended sediment. The similar R E E patterns, but widely varying absolute R E E abundances, for these suspended materials can be explained by variable dilution of phases with high REE concentrations and uniform R E E patterns (clay minerals) with phases of low R E E concentration such as quartz [18].

V

20

3.2. Dissolved loads

10

0.0

1.0

2.0 Nd

3.0

4.0

TOM(Ga)

Fig. 2. L a / Y b in suspended loads versus the Sm-Nd depleted mantle model ages (TffM d ) of these samples [15,16]. The degree of light REE enrichment appears to be directly related to the age of source rocks in each river drainage basin.

The NASC-normalized R E E patterns for river water dissolved matter are shown in Fig. 3. F r o m this figure it is apparent that these waters do not have a flat NASC-normalized R E E pattern. Lake Isua and the Great Whale River, two waters draining Archean gneisses, have light R E E enriched patterns with ( L a / Y b ) N = 3.3 and 3.1, respec-

39 10 /1

-< Z

~

-4

-4

. . . . . . . . . . . . . . .

10

-5

ou~

10

.

.

.

.

o o

~

~

v / V -

.

.

.

.

.

.

.

.

.

-o

o--

10

~ ° ~ "~'~-~-;-

° ~

.

0-5

-6

10

.

•

._~

w

9L

10

10

-7 '7 - - v 0hr* • - - • Pamponga o - - o $hInano

• - - • Pampanga -8

10

o - - o Shlnono I

I

I

Lo Ce

I

~

Nd

I

I

I

I

I

Sm Eu Gd

I

~

Dy

Er

I

I

I

Yb

10

,

. . . . . . . . .

r

. . . . . . . .

,

I

I

I

Nd

I

Sm

L

I

I

I

Gd

Eu

I

I

Dy

Er

I

I

Yb

I

ku

Fig. 4. REE abundances in dissolved loads normalized to the REE abundance of suspended material for each river. All rivers in this study have dissolved loads enriched in the heavy REE relative to suspended material. The most substantial heavy REE enrichment occurs in rivers of high pH. Negative Ce anomalies are apparent and are also most pronounced for alkaline rivers.

tively. The Pampanga and Shinano rivers from the Philippines and Japan have heavy REE enriched patterns, with ( L a / Y b ) N = 0.12 and 0.42, respectively. The Amazon, Indus, Mississippi, and Ohio rivers also have heavy REE enriched patterns in contrast to their suspended load patterns. Values of ( L a / Y b ) N for these rivers range from 0.38 for the A m a z o n to 0.14 for the Ohio River (Table 2). The Indus and Mississippi rivers have a pronounced negative Ce anomaly. Fig. 4 plots the REE patterns of the river water dissolved loads normalized to the suspended load . . . . . . . .

I

ko Ce

Fig. 3. REE abundances in dissolved loads normalized to shales (NASC). Major rivers have heavy REE enriched patterns relative to NASC, with (La/'Yb)r~ = 0.14-0.42.

1000

I

-8

Lu

values for each river. This type of normalization removes variations in relative R E E abundance that originate in the source rocks in a drainage area and allows REE fractionations caused by the weathering process to be examined. All of the rivers in this study have dissolved loads that are enriched in the heavy REE relative to the sus-

. . . . . . . .

•

(3

•

b

1O0 0 O []

10 V

V

,

,

J

,

,

n ,,]

,

,

,

,

C

•

o

•

10.0 O

•

0

0

D

o

OAmazon •

Great

Whole

~, I n d u s • Lake Isua

1.0

0 Mississippi

~' O h i o • Parnpanga O Shinano 0.1

.........

5

' ........................

6

7 pH

,,,,

,

,

,

,

8

,,

,,i

.

10 [Na

+ Ca]

.

.

.

.

.

.

.

100 (ppm)

Fig. 5. La and Yb concentrations in the dissolved load versus pH and (Na + Ca). The concentrations of the light REE are inversely related to pH and ionic strength. Heavy REE concentrations are less sensitive to these parameters.

40

pended load. A measure of this enrichment is given by: K dLa/Yb l - s l _-- ( L a / Y b ) d i s s / ( L a / Y b ) s u s

p

(2)

This ratio ranges from 1.0 for the Great Whale River and 0.73 for Lake Isua, to 0.56 and 0.43 for the Pampanga and Shinano rivers, to 0.17-0.065 for the Amazon, Indus, Mississippi, and Ohio rivers (Table 2). The degree of heavy REE enrichment appears to be greatest for rivers of high p H (Table 2). Ce anomalies are expressed as:

suspended load. Rivers of high p H ( = 8) have low light REE concentrations but are strongly enriched in the heavy REE. The Amazon River, with its low p H and heavy R E E enriched pattern, is an exception to these general statements. The concentration of any R E E in the < 0.2 ~ m load of rivers can be written as the sum of the REE concentration of very fine colloids, free ions in true solution, inorganic complexes of the REE, and organic complexes of the REE. Thus for La for example we have: [ta]

Ce* = 3 C e N / ( 2 L a N + N d N )

<0.2

~m

=

[ t a ] c l d + [ t a ] f r e e -4- [ta]inorg. cplx.

(3) + [La]org. cp,x.

where the subscript N refers to NASC normalized concentrations are given in Table 2. As shown, negative Ce anomalies (Ce* < 1) are present for 6 of the 8 waters studied and are most pronounced for rivers of high p H (Table 2). Large variations are also observed in the absolute abundances of the REE in the dissolved loads of the rivers in this study. For example, La concentrations range from 2.9 to 609 ppt, while Yb concentrations range from 0.94 to 15 ppt. In a study of the Sm-Nd systematics of about 40 river waters, we [8] found a general inverse relationship between log [Nd] and p H and a positive correlation between log [Nd] and log [Fe]. Because major cation concentrations and p H were positively correlated, a relationship between log [Nd] and log [Na + Ca] was established and used to estimate N d concentrations for river,4 with known [Na + Ca]. Fig. 5 plots [La] and [Yb] versus p H and [Na + Ca]. The plots for La are quite similar to those for Nd, indicating a strong p H control on the concentration of the light REE in solution. Yb concentrations are less sensitive to pH.

4. Discussion

4.1. Behavior of the R E E during, weathering and river transport The data presented above suggest that p H is a major factor in controlling both the absolute abundances of the light R E E in solution and the relative REE pattern of dissolved material. Dilute rivers of low p H have high R E E concentrations and REE patterns that are similar to those in the

(4)

The magnitude of each of these terms is dependent upon the REE under consideration, the type of rocks being weathered, and the chemistry of the solution. To evaluate this expression quantitatively, the stability constants for the REE with the major inorganic and organic anions must be known and the possible contribution from fine colloids must be established. There is a possibility that some of the R E E in the < 0.2/~m load of rivers are actually present as fine colloidal material. Concentrations of Fe and Mn in fresh waters are dependent upon the method of filtration: decreasing concentrations with decreased filter pore size are observed [19-22]. Stordal and Wasserburg [23] measured the concentration of Nd and Sm in the Amazon and Mississippi rivers. They found that concentrations were - 7 0 % lower in waters filtered to 0.1 btm relative to waters filtered to 0.4 ~tm. Most of the samples in this study were filtered to 0.2 #m, thus a significant percentage of the REE in the "dissolved" load of the samples in this study may not be in true solution. The use of the above information [23] to estimate the proportion of the REE associated with particles < 0.2 /zm in our samples requires knowledge of the size distribution of particles in rivers. If we assume that the concentration of particles (mass particle/mass water) in river water is constant for particles of 0.1 to 0.4/~m, then on average - 23% ( 1 / 3 of 70%) of the REE in the < 0.2 /zm load occurs in the colloidal state. The abundance of fine colloidal material in any particular river is probably sensitive to solution chemistry. Colloids destabilize and coagulate in solutions of high ionic strength, thus

41

dilute rivers can have a larger colloidal R E E contribution than alkaline rivers. Additional studies of the size distribution of colloidal R E E in rivers are needed to characterize the extent of colloidal transport of R E E in the < 0.2 /~m load of rivers. The concentration of free ion R E E in true solution is controlled by interactions with solid phases. Given the apparent systematic relationships between the REE in solution and suspended material, and the short residence time of water is rivers, it is most likely that the concentration of free ion R E E is controlled only by reactions with fast kinetics. This requirement strongly favors a mechanism such as adsorption or dissolution of R E E on particle surfaces. Laboratory experiments have shown that Ce adsorption on soils is strongly p H dependent for waters of p H < 7 [24], with Ce mobile at low p H but fixed on the clays at neutral or high pH. In addition, the R E E are known to co-precipitate with ferric hydroxide at a p H > 7. Thus free ion R E E concentrations are probably inversely related to pH. The relative abundance pattern of the free ion R E E is not known; however in the absence of any fractionation, it will be similar to that of suspended material. Ce is an exception, as its free ion concentration is also controlled by a redox equilibrium between the dissolved Ce 3+ and C e 4+ species. Ce depletion in rivers of high p H is probably a result of preferential removal of C e 4+ onto Fe-Mn oxide coatings of particles..This process is known to occur in the marine environment [25] and may also occur in high p H rivers. The abundances of the various inorganic complexes of the R E E can be calculated, as stabihty constants for the R E E and O H - , F - , CI-, SO 2 - , and CO 2- are available [26]. The heavy R E E form stronger complexes than the light R E E with such inorganic ions as CO 2- , O H , and F - [26]. Thus rivers of high p H with abundant CO32- and O H - are expected to be enriched in the heavy R E E relative to the source rocks in their drainage basins. This is illustrated in Fig. 6, where the calculated REE patterns of two typical river waters of p H 6 and 8 are plotted. In this diagram, it is assumed that the activities of the free trivalent R E E in solution are proportional to the R E E abundances in suspended material. Thus the variations in relative R E E abundance in this diagram are entirely caused by the formation of the com-

1000.0

O--OpH=6 100.0

F

,/

a~

/

10.0 w

1.0

0.1

0--0

i

i

La Ce

0

i

i

Nd

0--0~0--0

i

~

i

i

Sm Eu Gd

0

i

L

Dy

i

i

rr

0--0

i

i

i

Yb Lu

Fig. 6. Model calculation of total (free ion + complexed) REE concentrations in solution normalized to free ion REE 3+ concentrations for typical rivers of pH = 6 and 8. Inorganic complexing of the REE is minor for rivers of low pH, but is quite important for rivers of high pH. Complexing by CO32- and O H - is most pronounced for the heavy REE, resulting in heavy REE enriched patterns for alkaline rivers.

p l e x e s REE(OH) 2+, REE(SO4) +, and REE(CO3) +. For the river water of p H = 6, all of the R E E occur primarily as free ions in solution; thus one observes a relatively flat pattern. However, most of the REE in the river water of p H = 8 occur as carbonate complexes, and thus one observes enrichments in the heavy R E E such that k"dl-sl ~ La/Yb has a value of 0.05-0.1. It seems probable that formation of organic complexes of the R E E is important in rivers with abundant dissolved organic matter such as the Amazon; however no stability constants for R E E - h u m i c acid complexes are available to evaluate this effect quantitatively. It is difficult to explain these observations by means other than solution chemistry. For example, it is conceivable that the heavy R E E enrichments for river water dissolved loads could be a result of preferential weathering of easily soluble phases. The rivers that have large, heavy R E E enrichments drain regions of marine sediments, thus it is possible that preferential dissolution of a phase common to marine sediments with a heavy REE enrichment could produce the same enrichment in rivers. Unfortunately, there does not appear to be a phase with the required enrichment. C a r b o n a t e minerals [27,28] have values of ( L a / Y b ) N close to 1, as does fossil apatite [29]. Other minerals such as zircon and garnet which contain appreciable amounts of the heavy R E E are quite resistant to chemical weathering and could not be a source of a heavy R E E enriched solution. This it seems unlikely that the heavy

42 R E E e n r i c h m e n t of river water dissolved loads is due to preferential weathering of easily soluble phases. W e conclude that the inverse relationship between light R E E c o n c e n t r a t i o n s a n d p H m a y be due to p H - d e p e n d e n t destabilization of fine colloids or p H - d e p e n d e n t a d s o r p t i o n / d i s s o l u t i o n of R E E from particle surfaces. Ce depletions in some rivers of high p H are most likely caused by preferential scavenging of Ce 4+ o n t o particles. The enrichments in heavy R E E for rivers of high p H a n d low light R E E c o n c e n t r a t i o n are most easily exp l a i n e d b y the f o r m a t i o n of c a r b o n a t e a n d hydroxide complexes in solution.

4.2. Average R E E pattern of river water suspended material Because of the general u n i f o r m i t y of relative R E E a b u n d a n c e s of the 5 m a j o r rivers, these patterns m a y be representative of major rivers on a global scale. Previously published R E E patterns for the suspended load of the A m a z o n , Congo, Ganges, Mekong, a n d G i r o n d e rivers [1] are n o t very smooth, b u t do show a roughly similar overall pattern. Thus, from these data, we c a n reasonably estimate the relative R E E p a t t e r n of susp e n d e d matter in m a j o r rivers. The ratios L a / N d , C e / N d , S m / N d , etc. were d e t e r m i n e d by linear regression of the data for the five major rivers. The absolute p a t t e r n calculated, using an average N d c o n c e n t r a t i o n i n river water suspended m a t t e r of 36.4 p p m [15], is presented in Fig. 7. The average R E E p a t t e r n is light R E E enriched relative to N A S C , with ( L a / Y b ) N = 1:85 (Table 3). W e interpret this p a t t e r n as representative of the terrigenous i n p u t of the R E E from large cratonic

1.5

o --

o Average river

• --

• Average upper c r u s t ~ Average d e t r l f a l m a r i n e sediment

suspended load ~x- ~.~

~

0//°

1.o

~

~o 0.5

'

'

'

Lo Ce

~

'

Nd

'

~

~

S m Eu Gd

~

'

Dy

'

'

'

Er

o '

Yb Lu

Fig. 7. Shale (NASC) normalized R E E abundances of average

suspended material from major rivers, average upper crust from post-Archean shales [18], and average detrital oceanic sediments based on data given in [33-36]. Suspended material from major rivers and detrital oceanic sediments have heavy REE depleted patterns relative to both NASC and the Taylor and McLennan [8] average upper crust patterns.

l a n d masses to the oceans. However, a d d i t i o n a l a n d significant t r a n s p o r t of s u s p e n d e d s e d i m e n t to the oceans from island arcs [30] is expected to have flat or heavy R E E enriched patterns relative to N A S C , similar to the S h i n a n o a n d P a m p a n g a rivers in this study. C o n s e q u e n t l y , the average R E E p a t t e r n of the river s u s p e n d e d material flux to the oceans m a y be somewhat less heavy R E E depleted t h a n is suggested b y the R E E p a t t e r n s of the major rivers.

4.3. Average R E E pattern of river dissolved material A t t e m p t i n g to estimate a n average R E E p a t t e r n of river water dissolved material is a difficult endeavor given the large range i n absolute a n d relative a b u n d a n c e s of R E E observed in this study.

TABLE 3 Average river water and seawater concentrations and REE residence times (years) in seawater a

Suspended load Dissolved load Estuarine removal b Eft. diss. load c Seawater d Residence time

La

Ce

Nd

Sm

Eu

Gd

Dy

Er

Yb

Lu

(La/Yb) N

Ce *

39.6 30.8 0.27 8.3 5.3 21,000

80.9 64.5 0.40 25.8 1.8 2300

36.4 40.9 0.40 16.4 3.5 7100

6.91 10.8 0.46 4.97 0.65 4400

1.43 2.66 0.39 1.04 0.16 5100

5.33

4.18 11.5 0.47 5.41 1.2 7400

1.98 8.46 0.61 5.16 1.2 7800

1.68 6.06 0.57 3.45 1.2 12,000

0.25 ---

1.85 0.40 -0.19 0.33 -

1.03 0.91 -1.12 0.19 -

1.03 -

a Concentrations for river suspended load in ppm, for river water dissolved load and seawater in ppt. b Mass fraction of REE flux in rivers not removed in estuaries [38]. c Effective river water concentrations calculated from average river water and estuarine removal factors given in this table. 0 Average of 2.5 km data at 5 stations [5-7,39].

43 Additional data for the rivers of the Northwest U.S. [2] and the G a r o n n e - D o r d o g n e River in France [1] are also available for calculation of the average R E E pattern. This data base consists of rivers that have a total discharge of 7817 km3/yr, or roughly 18% of the total worldwide runoff. The Amazon River, with a discharge of 6930 km3/yr, is almost 90% of the total. As a consequence, any runoff weighted average of the available data will give an average R E E pattern which is essentially that of the Amazon River. Although the Amazon is the largest river in terms of discharge to the oceans, its chemistry is generally quite different from world average river water [9,31]. Consequently, the Amazone may be biased with respect to the REE pattern of global river runoff. An alternative method is to calculate an unweighted average for the REE in the 6 largest rivers (Amazon, Mississippi, Ohio, Columbia, Fraser, and Indus) in the data base. The result of this is given in Table 3. The unweighted average of major element concentrations for these rivers is slightly higher than the world average; however, the average Nd concentration of 41 ppt is almost equal to the value of 40 ppt obtained from a global balance [8]. The R E E pattern of this average is plotted in

10-5

I

I

]

I

10-4 b'l < Z ~-, 10-5

• __&JJ

&~.~_

~o~O--o~O----~

O

10-6 IM

• / 0 ~/

10-7

£~ ~'

10 - 8

I

La Ce

i

I

~ Nd

I

i

I

O--OAverage river water • - - • Effective river water flux to the oceonn • - - • Hydrotherrnol water A - - A Average ~eawaier I

5rn Eu Gd

I

I

Oy

I

I

Er

I

I

I

Yb

Fig. 8. NASC normalized R E E abundances of average river water dissolved material (Table 3), the effective river flux to the oceans (Table 3), average hydrothermal water [40,41], and average seawater [5-7,39]. The heavy R E E enrichment of average fiver water and the effective fiver flux to the oceans is similar to that of average seawater.

Fig. 8 and is strongly enriched in the heavy REE ( ( L a / Y b ) N = 0.4).

4. 4. Implications for the REE systematics of the continental surface Because of the very low concentration of the R E E in the dissolved load of rivers, the suspended load contains almost all of the R E E in rivers derived from weathering of the continental surface. Thus, the R E E pattern of suspended material can be used to estimate the REE pattern of the continental surface. In Fig. 7, we plot the average NASC-normalized R E E pattern of suspended material from the major rivers in this study and compare it to the average upper crust values of Taylor and McLennan [18]. The average values given in [18] were based on average post-Archean shale values. There is a significant discrepancy between these two patterns, with average suspended material depleted in the heavy R E E relative to the shale pattern ( ( L a / Y b ) N = 1.7). This difference between the R E E patterns for major river suspended loads and post-Archean shales has significant implications for the mass balance of particulate R E E in the oceanic environment. The data suggest that the source of the R E E in shales is not simply equal to the terrigenous material delivered by major rivers, and that an additional heavy R E E enriched source could be contributing to the shale REE pattern. It is possible that a part of the discrepancy between the shale and river suspended load REE patterns is caused by heavy mineral separation in rivers. Heavy minerals such as zircon and garnet have REE patterns that are heavy R E E enriched ( ( Y b / N d ) N = 50-190). During river transport, these minerals may be preferentially transported in the bed load relative to near surface waters. As a consequence, the samples in this study taken from the near surface of rivers m a y have a lower heavy mineral concentration than average crust. Measurements of H f concentrations in suspended loads can be used to determine whether this is reasonable, since H f is carried primarily in zircon. H f concentrations reported for the suspended load of 7 major rivers [32] range from 4.0 to 8.4 ppm. The typical value of H f in river suspended material of 6 p p m [32] is comparable to the concentration of H f in shales and upper crust (5-6 ppm; [18]). In addition, near shore detrital marine sediments

44 from the Zaire Fan [33] and the river bottom sample from the Amazon have light REE enriched patterns ( ( L a / Y b ) N = 1.0-2.0). These sediments presumably have a full complement of heavy minerals but nevertheless have a R E E pattern similar to the suspended particles in the rivers of this study. Thus, it does not seem likely that the discrepancy between shale and river water suspended load REE patterns in caused by differences in the concentration of heavy minerals. Additional measurements of REE patterns and Hf in the suspended and bed load of rivers are needed to verify this argument. It is also possible that the discrepancy is the result of differences in the age or composition of the rocks from which shales and river water suspended materials are derived. The REE pattern of river water suspended loads varies systematically with the Sm-Nd depleted mantle model age of these materials (Fig. 3), with L a / Y b increasing as T ~ increases. Thus the major fiver suspended loads may be sampling crust older than that sampled by post-Archean shales. However, the Tffd age of the North American Shale Composite (2.0 Ga; [14]) is greater than the T ~ age of suspended loads from the major rivers in this study (1.2-1.8 Ga), suggesting that differences in source rock age apparently are not the source of the discrepancy of shale and fiver water suspended material REE patterns. Source rock composition could also be a factor, since the REE pattern of silicic rocks is more light REE enriched than that of mafic rocks. However, comparison of the major element chemistry of the suspended material of major rivers [32] with that of average post-Archean shale [18] does not indicate any major differences. The difference between the REE patterns for river suspended loads and shales may be caused in part by changes in the nature of the continental crustal surface exposed to erosion through the Phanerozoic. This argument is supported by available REE data on modern shelf, slope, and abyssal sediments [33-36]. These data show that presentday detrital oceanic sediments are also depleted in the heavy REE relative to NASC ( ( L a / Y b ) N = 1.4, Fig. 7), although not to as large an extent as the major river suspended loads. Thus, we suggest that river suspended loads and recent detrital oceanic sediments more accurately reflect the composition of the modern continental surface than average

post-Archean shales. In particular, N A S C is a rather unusual mixture of sedimentary and metamorphic rocks from widespread geographic localities [12] and might not be as representative of the modern upper crust as major river suspended loads. We conclude that the REE patterns of major river suspended material or recent detrital oceanic sediments more accurately reflect the composition of the detrital input to the oceans than shales, and that the average values listed in Table 3 are more appropriate for normalization of modern continental erosion products than average postArchean shales.

4.5. Implications for the marine geochemistry of the REE Isotopic studies of Nd in seawater have shown that most of the dissolved REE in the oceans are derived, at least indirectly, from continental weathering [8,37]. Rivers are the primary means by which continental material is transported to the oceans. The concentrations, (7,.Rw, of the various R E E (i) in average river water (RW) dissolved material are presented in Table 3 and plotted in Fig. 8. This pattern is heavy REE enriched relative to NASC, with ( L a / Y b ) N = 0.40. This pattern is significantly affected by estuarine removal processes, however. It seems likely that R E E removal from solution in estuaries is common in rivers of both low and high pH. Data from the Gironde [1] and the Great Whale rivers [38] suggest REE removal of ~ 60% for the light REE and ~ 40% for heavy REE. Using the estuarine REE removal values (q~,) in Table 3, we calculated an effective concentration of the REE in the dissolved flux to the oceans (Ci Rw'eff= q~iciRW). Because of the somewhat greater removal of the light REE, the effective river flux is even more heavy REE enriched relative to NASC, with ( L a / Y b ) N = 0.19 (Fig. 8, Table 3). We calculated an average REE pattern for seawater (SW) based on measurements at 2500 m depth at five locations in the Atlantic and Pacific oceans [5-7,39] and the concentrations (7,sw are given in Table 3. The average pattern is enriched in the heavy REE ( ( L a / Y b ) N =0.33) and has a pronounced negative Ce anomaly (Ce* = 0.19).

45

The residence times of a species i in seawater with respect to the effective river water dissolved load flux is:

, r i [ C i sw

Msw)

(M.w,

(5)

Here the ratio of the ocean mass Msw = 1.39 x 10 24 g and the global river water discharge of )t;/RW = 4.2 × 1019 g / y r yield Msw/3J/RW = 33,000 years. Thus, since all the REE in seawater have lower concentrations than in the river input 1), their residence times in seawater must be less than 33,000 years. As shown in Fig. 8 La has an effective river concentration that is only slightly higher than that in seawater which results in a residence time of 22,000 years (Table 3). The largest concentration difference between river and seawater is observed for Ce and the values listed in Table 3 imply a residence time for Ce in seawater of 2300 years. Estimates of the seawater residence times of the other REE given in Table 3 range from 4400 years for Sm to 12,000 years for Yb. For N d in particular we obtain a seawater residence time of 7100 years. Using the average concentration of the REE in hydrothermal waters at 1 3 ° N and 2 1 ° N East Pacific Rise [40,41] (Fig. 8) and a hydrothermal water flux of 2.9 × 1016 g / y r [8], we estimate that the hydrothermal water/effective river water flux ratios for La, Nd, and Yb are 0.115, 0.026, and 0.007, respectively. The value of the flux ratio for Eu is 0.28, which reflects the large positive Eu anomaly observed in hydrothermal waters. If the rate of removal of Eu from seawater is similar to that of adjacent trivalent REE, then this positive Eu anomaly should be observed in seawater. However, the absence of an Eu anomaly in seawater suggests the presence of additional REE sources which would lower the relative contribution of Eu from hydrothermal waters, or it may indicate that the hydrothermal Eu flux is substantially lower than the value suggested by the available data. The presence of additional sources of REE in seawater can also be suggested by the residence times calculated above. These residence times are significantly longer than the currently accepted value for the mixing time between surface and deep water of = 1000 years based on 14C [42]. Piepgras and Wasserburg [37] used the differences

(ciSW/fiRW'efe<

in 143Nd/144Nd between various oceans and a simple steady-state two-box model to place constraints on the residence time of N d in the oceans. They concluded that the residence time of N d in the Pacific Ocean was ~< (404 years)/q where q is the entrainment efficiency for mixing Pacific Water into the Atlantic via the Antarctic Circumpolar Current. The long residence time obtained by us indicates a very low value for q of - 0 . 0 6 . Recently, Piepgras and Wasserburg [44] observed that 143Nd/144Nd variations in the North Atlantic are strongly coupled with major water masses, suggesting that both the water masses and Nd have comparable lifetimes for mixing and N d precipitation. They estimated that the residence time of Nd in seawater must be greater than 1000 years. The Nd residence time of 7100 years obtained by us could suggest that there are additional inputs of the REE to the oceans that are greater than the effective river flux, or that there are problems in our definition of "dissolved" REE in natural waters. A large aeolian flux of REE derived from atmospheric dust transport from the Sahara Desert has been suggested on the basis of surface water enrichments in the eastern Atlantic Ocean [5]. However, N d enrichments and low end values characteristic of wind-blown material from continental Asia are not observed in Pacific surface waters. Thus, dissolution of airborne material may not be of global significance. The diffusion of REE from oceanic sediments is also a potential source of REE in seawater. A progressive increase with depth in the concentration of REE in pore waters of shelf sediments from Buzzards Bay in the Northwest Atlantic Ocean has been observed [43]. This suggests that REE concentrations in pore waters from reducing environments may be substantially greater than seawater, and that the associated global fluxes of REE may be significant. Such sources would have a "continental" N d isotopic composition, and would thus be generally compatible with N d isotopic data in seawater. It is also possible that a fraction of the REE in seawater may be present as fine colloidal material ( < 0.4/~m) rather than in true solution. In such a situation, the river source of this material would be significantly greater than the estimates presented above, and REE residence times would be

46

smaller. A better description of the "dissolved" component of the R E E in natural waters is required to test this hypothesis. Future studies should address questions relating to the speciation and size distribution of "dissolved" REE in seawater and river water.

drology, Pakistan, for providing samples. We also wish to thank Henry Shaw and two anonymous reviewers for their comments on an earlier version of this paper. This research was supported by N S F grants E A R 82-06954, EAR 85-11912, and N A S A grant NAG9-90.

5. Conclusions

References

Neither river water dissolved nor suspended materials from the major rivers in this study have flat REE patterns relative to shales. The REE pattern of the suspended loads of small rivers is dependent on drainage basin geology. However, major river suspended loads have relatively uniform REE patterns and are heavy REE depleted relative to NASC. Major river dissolved loads have marked relative heavy REE enrichments relative to NASC and suspended material. These data suggest that the average detrital REE flux from rivers to the oceans is heavy REE depleted relative to shales, whereas the average "dissolved" REE flux to the oceans is heavy REE enriched. The available data from the suspended loads of major rivers and detrital oceanic sediments suggest that the modern continental surface is heavy R E E depleted relative to shales. Consequently, we suggest that the average REE patterns of major fiver suspended loads or detrital oceanic sediments are more appropriate for normalization of modern erosion products than the NASC R E E abundances. The new fiver water dissolved load data leave questions regarding the possible deficit of "dissolved" REE in the oceans unresolved. Calculated residence times for the dissolved REE in seawater with respect to the effective river input of dissolved REE range from 2300 to 21,000 years, are smallest for Ce, and greatest for the heavy REE and La. These large residence times might suggest the presence of major additional, yet unidentified sources of dissolved REE in seawater. More data on potential REE sources and a better understanding of the nature of "dissolved" REE in natural waters are needed to resolve this problem.

Acknowledgements We thank Steve Dobos, John Edmond, Lew Fox, and the Director of Surface Water Hy-

1 J.M. Martin, O. Hogdahl and J.C. Philippot, Rare earth element supply to the oceans, J. Geophys. Res. 81, 3119-3124, 1976. 2 K.M. Keasler and W.D. Loveland, Rare earth elemental concentrations in some Pacific Northwest rivers, Earth Planet. Sci. Lett. 61, 68-72, 1982. 3 J. Hoyle, H. Elderfield, A. Gledhill and M. Greaves, The behavior of the rare earth elements during mixing of river and sea waters, Geochim. Cosmochim. Acta 48, 143-150, 1984. 4 D.Z. Piper, Rare earth elements in the sedimentary cycle: a summary, Chem. Geol. 14, 285-304, 1974. 5 H. Elderfield and M.J. Greaves, The rare earth elements in seawater, Nature 296, 214-219, 1982. 6 H.J.W. de Baar, M.P. Bacon and P.G. Brewer, Rare-earth distributions with a positive Ce anomaly in the Western North Atlantic Ocean, Nature 301, 324-327, 1983. 7 H.J.W. de Baar, M.P. Bacon, P.G. Brewer and K.W. Bruland, Rare earth elements in the Pacific and Atlantic Oceans, Geochim. Cosmochim. Acta 49, 1943-1959, 1985. 8 S.J. Goldstein and S.B. Jacobsen, The N d and Sr isotopic systematics of river water dissolved material: implications for the sources of N d and Sr in seawater, Chem. Geol. (Isot. Geosci. Sect.) 66, 245-272, 1987. 9 R.F. Stallard and J.M. Edmond, Geochemistry of the Amazon, 2. The influence of geology and weathering environment on the dissolved load, J. Geophys. Res. 88C, 9671-9688, 1983. 10 J.G. Crock, F.E. Lichte and T.R. Wildeman, The group separation of the rare-earth elements and yttrium from geologic materials by cation-exchange chromatography, Chem. Geol. 45, 149-163, 1984. 11 M.F. Thirwall, A triple filament method for rapid and precise analysis of rare earth elements by isotope dilution, Chem. Geol. 35, 155-166, 1982. 12 L.P. Gromet, R.F. Dymek, L.A. Haskin and R.L. Korotev, The " N o r t h American shale composite": its compilation, major and trace element characteristics, Geochim. Cosmochim. Acta 48, 2469-2482, 1984. 13 E.S. Gladney, C.E. Burns and I. Roelandts, 1982 compilation of elemental concentrations in eleven United States Geological Survey rock standards, Geostandards Newsl. 7, 3-226, 1983. 14 M.T. McCulloch and G.J. Wasserburg, Sm-Nd and Rb-Sr chronology of continental crust information, Science 200, 1003-1011, 1978. 15 S.J. Goldstein and S.B. Jacobsen, The N d and Sr isotopic systematics of river water suspended material: implications for crustal evolution, Earth Planet. Sci. Lett. 87, 249-265, 1988.

47 16 S.L. Goldstein, R.K. O ' N i o n s and P.J. Hamilton, A Sm-Nd isotopic study of atmospheric dusts and particulates from major river systems, Earth Planet. Sci. Lett. 70, 221-236, 1984. 17 V.I. Korzun, ed., World Water Balance and Water Resources of the Earth, Studies and Reports in Hydrology, 663 pp., U N E S C O 25, Paris, 1978. 18 S.R. Taylor and S.M. McLennan, The Continental Crust: Its Composition and Evolution, 312 pp., BlackweU, Oxford, 1985. 19 V.C. Kennedy, G.W. Zellweger and B.F. Jones, Filter pore-size effects on the analysis of A1, Fe, Mn, and Ti in water, Water Res. 10, 785-790, 1974. 20 G. Figueres, J.M. Martin and M. Meybeck, Iron behavior in the Zaire estuary, Neth. J. Sea Res. 12, 329-337, 1978. 21 R.M. Moore, J.D. Burton, P.J. LeB. Williams and M.L. Young, The bebaviour of dissolved organic material, iron, and manganese in estuarine mixing, Geochim. Cosmochim. Acta 43, 919-926, 1979. 22 D.P.H. Laxen, W. Davison and C. Woof, Manganese chemistry in rivers and streams, Geochim. Cosmochim. Acta 48, 2107-2111, 1984. 23 M.C. Stordal and G.J. Wasserburg, N e o d y m i u m isotopic study of Baffin Bay water: sources of REE from very old terranes, Earth Planet. Sci. Lett. 77, 259-272, 1986. 24 R.E. Brown, H.M. Parker and J.M. Smith, Disposal of liquid wastes to the ground, in: U.N. Internal Conference on the Peaceful Uses of Atomic Energy, 9, pp. 669-675, United Nations, New York, N.Y., 1955. 25 E.D. Goldberg, M. Koide, R.A. Schmitt and R.H. Smith, Rare earth element distribution in the marine environment, J. Geophys. Res. 68, 4209-4217, 1963. 26 D.R. Turner, M. Whitfield and A.G. Dickson, The equilibrium speciation of dissolved components in freshwater and seawater at 25 ° C and 1 atm pressure, Geochim. Cosmochim. Acta 45, 855-881, 1981. 27 P.P. Parekh, P. Moiler, P. Dulski and W.M. Bausch, Distribution of trace elements between carbonate and noncarbonate phases of limestone, Earth Planet. Sci. Lett. 34, 39-50, 1977. 28 M.R. Palmer, Rare elements in foraminifera tests, Earth Planet. Sci. Lett. 73, 285-298, 1985. 29 J. Wright, R.S. Seymour and H.F. Shaw, REE and N d isotopes in conodont apatite: Variations with geological age and depositional environment, Geol. Soc. Am. Spec. Pap. 196, 325-340, 1984. 30 J.D. Milliman and R.H. Meade, World-wide delivery of river sediment to the oceans, J. Geol. 91, 1-21, 1983.

31 M. Meybeck, Concentrations des eaux fluviales en 616ments majeur et apports en solution aux oceans, Rrv. Grol. Dyn. Grogr. Phys. 21,215-246, 1979. 32 J.M. Martin and M. Meybeck, Elemental mass-balance of material carried by major world rivers, Mar. Chem. 7, 173-206, 1979. 33 T.C.E. van Weering and G.Th. Klaver, Trace element fractionation and distribution in turbidites, homogeneous and pelagic deposits; the Zaire Fan, Southeast Atlantic Ocean, Geo-Mar. Lett. 5, 165-170, 1985. 34 J. Thomson, M.S.N. Carpenter, S. Colley, T.R.S. Wilson, H. Elderfield and H. Kennedy, Metal accumulation rates in northwest Atlantic pelagic sediments, Geochim. Cosmochim. Acta 48, 1935-1948, 1984. 35 W.M. White, B. Dupr6 and P. Vidal, Isotope and trace element geochemistry of sediments from the Barbados Ridge-Demerara Plain region, Atlantic Ocean, Geochim. Cosmochim. Acta 49, 1875-1886, 1985. 36 E.R. Sholkovitz, Rare earth elements in the sediments of the North Atlantic Ocean, A m a z o n delta, and East China Sea: reinterpretation of terrigenous input patterns to the oceans, Am. J. Sci., in press, 1988. 37 D.J. Piepgras and G.J. Wasserburg, N d isotopic variations in seawater, Earth Planet. Sci. Lett. 50, 128-138, 1980. 38 S.J. Goldstein and S.B. Jacobsen, REE in the Great Whale River estuary, northwest Quebec, Earth Planet. Sci. Lett. 88, 241-252, 1988. 39 G. Klinkhammer, H. Elderfield and A. Hudson, Rare earth elements in seawater near hydrothermal vents, Nature 305, 185-188, 1983. 40 A. Michard, F. Albarede, G. Michard, G.F. Minster and J.L. Charlou, Rare-earth elements and u r a n i u m in hightemperature solutions from East Pacific Rise hydrothermal vent field (13 ° N), Nature 303, 795-797, 1983. 41 A. Michard and F. Albar~de, The REE content of some hydrothermal fluids, Chem. Geol. 55, 51-60, 1986. 42 W.S. Broecker and T.H. Pengo Tracers in the Sea, 690 pp., Eldigio Press, Lamont-Doherty Geological Observatory, Palisades, New York, N.Y., 1982. 43 H. Elderfield and E.R. Sholkovitz, Rare earth elements in the pore waters of a coastal sediment, EOS Trans. Am. Geophys. U n i o n G5, 956, 1984. 44 D.J. Piepgras and G.J. Wasserburg, Rare earth element transport in the western North Atlantic inferred from N d isotopic observations, Geochim. Cosmochim. Acta 51, 1257-1271, 1987.