Authigenic Uranium in Marine Sediments of the Benguela Current Upwelling Region During the Last Glacial Period

Anomalously high uranium contents and Z3SUp32Th activity ratios de posited during the oxygen isotope stages 2 to 4 are observed in the MD962085 core from the Benguela Current upwelling area. In conjunction with the activity ratio of 234UfZ3 8 U, the high uranium contents found in this core can not be considered as detrital; instead, they must be of authigenic origin. The high biological productivity in the overlying seawater may have in~ duced a high flux of organic matter resulting, directly or indirectly, in a reducing environment, and may have led to the addition of authigenic ura nium to the sediments during the last glacial period. The correspondence between the variations in authigenic U content, % TOC, and 230Th -nor~ ex malized TOC flux supports this suggestion. The observed high paleoproductivity during the last glacial period of the core may be due to the greater intensity of upwelling in the Benguela Current upwelling sys tem.


INTRODUCTION
Th and U isotopes, and TOC content in a gravit y core were measured to examine the diagenetic behavior of uranium in the Benguela Current upwelling region. The core was collected by R/VMarion Dufresne during the IMAGES (International Marine Past Global Changes Study Pro gram) II cruise.
Uranium and its daughters have been shown to be valuable as geochronometers (e.g. Broecker and Peng, 1982;Bard et al., 1990) and paleoceanographic proxies (e.g. Francois et al., 1993;Yu et al., 1996), resulting in intensive study of the marine geochemistry of uranium for decades (Cochran and Krishnaswami, 1980;Anderson, 1982;Colley and Thomson, 1985;Cochran et al., 1986;Wallace et al., 1988;Anderson et al., 1989;Bames and Cochran, 1990;Klinkhammer and Palmer, 1991). These studies have suggested that uranium behaves conservatively with the formation of soluble U(VI) carbonate complexes in oxygenated seawater; however, under reducing conditions soluble U(VI) in the overlying seawater can be transformed into insoluble U(IV). This transformation, whieh can occur in the seawater column, at the sediment-water interface, and within the sediments, resulted in the addition of authigenic uranium to sediments, and altered the uranium distribution in the sedimentary records.
It became clear that uranium is reactive in sorne marine environments such as anoxic basins (e.g. Anderson, 1989), coastal oceans (e.g. Cochran et al., 1986), turbidites (e.g. Colley and Thomson, 1985) and pelagie sediments (e.g. Wallace et al., 1988). For sediments in upwelling areas, a relatively high flux of partieulate organic matter can easily make the regions suboxie or anoxie environments; uranium is thus easily able to behave unconservatively. This paper adds detail to CUITent knowledge on the behavior of dissolved uranium in sediments of the Benguela CUITent upwelling region, and also examines how the uranium distribution in the sedimentary records corresponds to the variations of the ocean conditions in the region with the glacial to interglacial dimate changes.
The surface water layer at the core site is strongly influenced by the Benguela CUITent, which flows northward along the southwest African coast mainly between 34 ·S and 15 ·S (Nelson and Hutchings, 1983). This northward flow begins to bend northwest and flows away from the coast at about latitude 30 oS (Nelson and Hutchings, 1983;Stramma and Peterson, 1989; Figure 1). To the west, this part of the CUITent is juxtaposed to the coastal CUITent from the geostrophic CUITent of the South Atlantic antieydonie gyre (Stramma and Peterson, 1989). Due to the prevailing wind, the coastal upwelling develops within the eastemmost section of the Benguela CUITent, whieh has been referred to as the Benguela CUITent upwelling system (Shannon, 1985).
Paleontologieal and sedimentological studies have shown that the upwelling activity of the Benguela CUITent region has existed for millions of yearS (Siesser, 1978;Diester-Haass, 1985;Oberhansli, 1991). The intensity and distribution of Benguela CUITent upwelling system are important not only to surface area of the high productivity zone and in attendant fluxes of sedimentary materials to the seafloor, but also to the amount of upwelled cool, nutrient-rlch and 02-depleted waters. AU of these variations are important factors which affect the diagenetic behavior of uranium in the sediments. .

Sample Material
The MD962085 core was sampled at 29°42'8, 12°56'E in a waterdepth of 3001 m ( Figure  1). The site is located off the coast of the Orange River where the surface water layer is strongly influeneed by the Benguela CUITent upwelling system. The core is 35.4 m in length, but only the first 4 meters has been analyzed and discussed. The data represent the whole last glacial period from oxygen isotopie stages 2 to 4 ta the present time. Ta provide a highresolution record, most of the core was sampled at intervals of 5-10 cm, whieh represented a frequency of approximately 1 to 2 kyr.

203
Analysis of Th and U isotopes wAll samples were subjected ta total dissolution with mixtures ofHN0 3 , HF, and HCI0 4 in the presence of 236U and 229Th as yield monitors. The radiochemical analyses followed the method of Anderson and Fleer (1982) with sorne modifications. Briefly, Th isotopes were separated from U isotopes by elution with 9N HCI from AGl-X8 anion exchange resin; after Th elution, U isotopes were eluted with dilute HCL Both Th and U solutions were evaporated to near dryness, put into 8N HNO" and purified separately by using an 8N HN0 3 anion exchange column. The recovered Th and U solutions were evapotated separately to a small drop of concentrated HN0 3 and taken up in O.OlM HN0 3 and 2M NH 4 Cl hefore electroplating. Th and U isotopes were determined by counting emitted alpha-partic1es with silicon surface-barrier detectors. The analytical precision (1 0) is reported in Table 1.
The final uncertainties of the radioisotope analyses in Table 1 are from several sources. The primary contribution is from the number of raw counts of spectra for individual isotopes. Other factors, such as background and decay correction and low sample weight, also affect.the final uncertainties. Additional uncertainties for calculating radioactivity ratio are included the propagation of all sources of errors, which is discussed in detail by Ku (1966). The activities of U and Th isotopes are ca1culated from the raw counts for the isotope in a given peak area and are subtracted from the number of background counts. For 238U, 236U, and 234U, the peak area of each isotope inc1udes the same number of channels because the shape of the spectra for all U isotopes. For 232Th, 23% and 229Th, the shapes of the spectra are different for each isotopes. The number of counts of the two channels on the two sides of each Th peak area was taken account for 0.5% of the total integrated counts in the peak area. The background correction was made taking the number of the background counts to be equal to that in the same area (with the same number of channels) of each peak area of the isotope. For 23CTh and 229Th isotopes, the largest sources of background were generally from low energy tailing counts of higher energy isotopes. Therefore, tail correction was combined with background correction. For more details, refer to Yu (1994).
Organic Carbon-Total carbon and organic carbon content were determined using a HORIBA carbonlsulfur analyzer. The analytical procedure from Hedges and Stern (1984) and Chang et al. (1991) was followed. A preweighed sample of 0.07-0.1 g was first analyzed for total carbon content using the analyzer, and was then treated with fuming HCI to leach out the inorganic carbon and thus allow the analysis of organic carbon. After the fuming HCI acidification, the carbon content determined is defined as the content of organic carbon. The fuming HCI acidification was suggested as the best method to remove carbonate from marine sediments (Chang et al., 1991). Replicate measurements of a standard for total carbon analyses give an average value of 0.673±O.005 %C (10', n=5) with an accuracy of ± 0.03%. The blanks for total carbon ranged from 0.002 to 0.004%C.
AMS He dating-The monospecific samples of G. inflata of several sections of the core were used to generate the AMS 14C age. The monospeeific samples of G. inflata were handpicked from the dry coarse fraction, >63 mm size. Foraminifera were ultrasonically c1eaned using methanol to remove adhering partic1es, followed by soaking in a solution of 5% NaCIO for 12 hours. The c1eaned foraminifera samples were converted into CO 2 gases and collected via vacuum line in the stable isotope laboratory of the Institute of Barth Sciences, Academia Sinica, Taipei. The radiocarbon measurements of the CO 2 gas samples were made by the Rafter Radiocarbon Laboratory at the Institute of Geological and Nuc1ear Sciences in New Zealand. The AMS 14C dates are calculated conventionally (i.e., relative to 0.95 National Bureau of Standards oxalic acid using the Libby half-life of 5568 years). The data of core MD962085 are available electronically at Paleoceanographic Data Center of Core Laboratory-Center for Ocean Research, NSC, at the Institute of Applied Geophysics, National Taiwan Ocean University, Keelung, Taiwan, R.O.C. (lntemet:http://140.121.175.1l4).

Age Mode} and Sedimentation Rates
Age model for the MD962085 core was established by both radiocarbon dating and oxygen isotope stratigraphy. No correction was made for the difference in 14Cp2C ratios between surface ocean :r. CO 2 and atmospheric CO 2 for the AMS t4C ages. This correction is currently about 400 years for most places in the surface ocean (Broecker et al., 1988). However, the AMS 14C ages established for MD962085 were converted into caIendar ages based on CALIB 3.03C (Stuiver and Reimer, 1993), a pro gram used to calibrate 14C ages with respective to treerings. Table 1 lists the calculated model ages for the top 110 cm ofthe core which is developed on the basis of linear interpolation between adjacent calibrated radiocarbon dates.
Since no AMS 14C data are available below the top 110 cm, oxygen isotope was employed as the stratigraphie framework for deeper sections of the core. The high resolution oxygen isotope record of the core was established by Chang (1997). The same planktonic foraminifera G. inflata was aIso utilized for oxygen isotope measurement because of the abundance of this species in the region (B'e, 1977). The oxygen isotope record of the core was then compared with the orbitally-tuned, stacked, standard oxygen isotope record from Martinson et ai. (1987), and control points of the oxygen isotope record were thus identified. Precise correlation with the current data base proved difficult, but the comparison enabled identification of the boundaries of isotope stages 112,2/3, and 3/4 at 73.5 cm, 136 cm, and 348.5 cm, respectively. The control points were then used to establish the stratigraphic framework below 110 cm of the core.
Sedimentation rates of the core were ca1culated on the basis of the calibrated AMS 14C dates. The ca1culated sedimentation rates are 7.8 cmlkyr for the Holocene period (3.5 cm to 73.5 cm), and 7.4 cmlkyr for the last glacial maximum (73.5 cm to 108.5 cm). Based on the oxygen isotope stratigraphy (Chang, 1997), the average sedimentation rate of the core from the top down to 26 m is approximately 5 cmlkyr.

Radioisotope Records
The depth profiles of mU and 232Th concentrations for the MD962085 core are presented in Figure 2. A low concentration of 23BU (ranging from -0.3 to 0.6 dpmlg) was found in the uppermost sediments, while remarkably high 238U concentrations (ranging from > 1.9 to -4.5 dpmlg) were found throughout the rest of the core between 73.5 and 400 cm (Table 1, Figure  2). According to AMS t4C age and oxygen isotope stratigraphy, this sharp increase in U concentration corresponds to the boundary between oxygen isotope stages 1 and 2. Then, these values remained high throughout the last glacial period.
In contrast to U 238 concentrations, 232Th exhibited low activities ranging from -0.4 to 0.8 dpmlg throughout the core. 232Th is one of the most particle-reactive elements, and is the only non-radiogenic isotope of Th. In the ocean, 232Th is essentially locked in the lattice structure of minerais, and derived entire1y from continents through eolian and riverine pathways. Therefore, the low concentrations of 232Th in the MD962085 core indicate a small input of lithogenic material into the deep-sea sediments where it located. The relatively low contents of terrigenous material (ca1culated from the residuaI fraction of total sedimentary components subtracting the %opal+%CaC0 3 +%TOC; Yu, unpublished data) in this core are consistent with this explanation. Although 232Th activities in the core are low, they still displayed variations corresponding to the glacial-interglacial c1imate change (Table 1, Figure 2). These variations are due to enhanced terrigenous material input from the Orange River and greater input of  The vertica1lines delineate the core sections corresponding to oxygen isotope stage 2, 3, and 4, respectively. 207 eolian dust during the low stands of sea level and drier condition during the glacial period .

Authigenic,U Profile
High 238V and low 232Th concentrations of samples during the last glacial period in the MD962085 core indicate that both authigenic and lithogenic V existed in the sediments. Since only total V content ofthe sediments can be measured, it is necessary to derive the authigenic U contents of these samp1es by an indirect means. Virtually aIl of the 23ZTh in marine sediments are lithogenic, and typical 238Up32Th activity ratios for major crustal rocks and pelagie marine sediments are in the range ofO.8±O.2 (Anderson, 1982;Anderson et al., 1989;Wedepohl, 1995). Therefore, detrital 238 V can be estimated from 232Th activities, and authigenic U content of sediments is calculated by the difference between measured total 238V activity and detrital 238V activity as where 238V and 232Th are the measured 238V and 232Th content respectively, and (0.8±O.2) * m m 232Th m represents the estimation of the mean activity of detrital 238U in dpmlg (Anderson et al., 1989). The results, displayed as % authigenic V, are shown in Figure 3a. Clearly, high activity ratio of 238Vp32Th existed between 73.5 and 400 cm of the core (Table 1), and a large fraction of the uranium (Figure 3a) found during the whole last glacial period in this core is not detrital; rather, the addition al U is authigenic.

Implications for the Authigenic Uranium Enrichment in the Glacial Sediments
Enhanced concentrations of uranium in deep-sea sediments have been recognized on hydrothermally active mid-ocean ridges (MORs; Fisher and Bostrom, 1969;Bender et al., 1971), andalso in reduced sediments (Cochran and Krishnaswami, 1980~ Bames and Cochran, 1990~ Klinkhammer and Palmer, 1991. Due to the remoteness of this core from any hydrothermal inputs, it is unlikely that hydrothennal activity is the causative factor of U enrichments observed in MD962085. Rather, additional U observed in the core may have been added to the sediments as a result of reducing conditions either in the sediment deposited during last glacial period_or within the water column. Generally, the large fraction of authigenic U must have been added to the sediment as a result of two processes: (1) biogenic fixation of U in the water column and scavenging of U by microorganlsms, or precipitation of U from solution as discrete U minerais or adsorption as a hydrolyzed species to the surface of existing mineraIs or by reducing conditions in micro environments within settling particles (Fisher et al., 1987;Anderson, 1982;Anderson et al., 1989) and (2) U precipitation at the water-sediment interface where higher proportions of reactive organic matter may enhance bacterial activity, which acts as a trap for uranium either directly (Mann and Fyfe, 1984) or indirectly by lowering the redox potential during the early diagensis of organie matter (Cochran and Krishnaswami, 1980;Bames and Cochran, 1990;Klinkhammer and Palmer, 1991). Indeed, both processes are major mechanisms of V incorporation into anoxie organic-rich marine sediments. Wh en reducing conditions occur close to the sediment-water interface, additional U is added to the sediments from the water column through the steps of the reduction of the soluble VeVI) in the overlying sea water, the removal of the soluble V (VI) from the bottom water or the pore water of the sediments, and precipitation of the insoluble V(IV) as the solid phase uranium in the sediment. A pore water gradient is thus created from the sediment-water interface down to the deeper depths of the sediment (Cochran and Krishnaswami, 1980;Bames and Cochran, 1990;Klinkhammer and Palmer, 1991), and the removal of uranium into the solid phase thus affects the sedimentary record. This removal process leads marine sediments to a substantiai sink for the uranium dissolved in seawater. Thé authigenic uranium added to the sediments by this process accumulates at a rate which is established by pore water U concentration gradient, which depends on the level of oxygen in the bottom water and the flux of metabolizable organic matter from both export production, whieh originates from the water above, and accumulation rate which is supplied laterally (Le., sediment focusing). Lacking pore water V profile and other data to allow further investigation on the variation in level of oxygen in bottom water, in the following we attempt only to examine how the variations in flux of organic matter account for addition al U uptake in the sediments deposited during the last glacial period of the MD962085 core. %TOC of the core is analyzed, and used toca1culate 23O'fh eK -normalized TOC flux to represent the relative supplies of organic matter from the overlying sea water in the region where the core is collected.
From the results, the variations in % authigenic U (Figure 3a) are observed to correspond weIl to the variations in %TOC of the MD962085 core (Figure 3b). The sharp increase in authigenic U at circa 12 kyr BP also coincides with the sharp increase in %TOC. Values for both authigenic U and %TOC remain high for the last full glacial period, Le .. stages 2 to 4. In addition, the variations in 23orhex~normalized TOC flux were consistent with those in both %TOC and authigenic U (Figure 3c). 230Th ~normalized TOC flux was calculated by normalizing TOC flux to 23D'fh activity ~ ~ in sediments. 230Thex-normalized flux method was first proposed by Bacon (1984), and has been discussed extensively elsewhere (Suman and Bacon, 1989;Francois et aL, 1990). Since frrst proposed, the method has been applied heavily in paleoceanographic studies (e.g., Francois et al., 1990;Kumaret al., 1993;Yu, 1994Yu, , 1996. This method is based on observations from sediment trap studies suggesting that annually averaged fluxes of excess 23D'fh cnD'fh ) to ex the seafloor are nearly constant and close to the expected rates of production from the radioac~ tivity decay of 234U dissolved in the overlying water column (Yu, 1994). Thus,23D'fh can be ex used as a reference against which the flux (F) of other sedimentary components can be esti~ mated: where p is the production rate of 2311'J'h in the water column, z is the water depth, fi is the weight fraction of component i in the sediment, and 230J'h ex 0 is the activity of excess 23D'fh (Le. ,scavenging 230Th) decay corrected to time of deposition using an independent time scale based on 1 4 C or 8 18 0 stratigraphy. By normalizing to 230Th ex 0, each measurement gives a flux estimate at each sampled point and thus allows better time resolution. It has been proven that the 230rhex -normalized paleoflux method can correct the influence of post-depositional sediment redistribution by bottom current or sediment winnowing (Suman et al., 1989;Francois, et al., 1990). Thus, it can provide better estimates of preserved particle rain rate on the sea floor (e.g., Suman et al., 1989;Francois, et al., 1990;Kumar et al., 1993).
The higher levels of both %TOC and 230Th -normalized TOC flux in the same sequence ex with depth of the MD962085 core indicate higher paleoproductivity during the last full glacial period, including oxygen isotope stage 2, 3 and 4, than that during the Holocene in the study region (Figure 3b and c). These observations support the interpretation of high authigenic U may have added to the sediment as a result of more U fixation in or association with biogenic materials is scavenged. Alternatively, the dissolved U is added to sediments in its diagenetic sequence when more labile organic carbon is consumed at the sediment-water interface. The decomposition of more organic matter caused higher oxygen consumption, and led to the development of reducing conditions being closer to the sediment water interface. More labile organic carbon in the sediment is produced by a higher settling flux of biogenic materials from the overlying water column. The higher flux of the biogenic materials may result from more intensive upwelling of the Benguela upwelling system during the glacial periods. Intensive upwelling brought deep nutrient-rich waters to the ocean surface, and provided more nutrient for the planktonic communities in the surface water. The higher paleoproductivity during the last glacial period in this core is also suggested by Chang (1997) from his data on the higher abundance of planktonic foraminifera. In deep-sea sediments, the roles of "organic activity" in the water column and of early diagensis of organic carbon are difficult to weigh (Hillaire-Marcel et al., 1990). However, the uranium dissolved in sediment pore water is often characterized by high 234U/238U ratios (as high as 1.23±0.03; Cochran and Krishnaswami, 1980) compared ta those for uranium dissolved in the water column. Uranium dissolved in seawater has a typical 234 U/23SU activity ratio of 1.14 (Chen et al., 1986) whereas weathered lithogenic material often has a ratio < 1.0 due ta preferentialleaching of the radiogenic 234U. Thus, the relative1y high 234Up38U activity ratios with an average of 1.1O±O.04 (1 0' for 30 analytical data) measured in the sediment sections over the last glacial period of this core (Table 1) support the explanation of "organic activity" in the water column. However, more accurate measurement of the 234Up38U activity ratio by mass spectrometry is needed to elucidate the argument.
A similar pattern, with high 238Up32Th activity ratios in sediment deposited during the high productivity period, has been observed off the present-day NW African upwelling area (Mangini and Diester-Haass, 1983), in the Labrador Sea (Hillaire-Marcel et al., 1990), and in the north of the Polar Frontal Zone (Francois et al., 1993). High authigenic U from the last glacial stage in the sediments from the northeast Atlantic Ocean was also found, but it was interpreted differently. Cores from Cape Verde Rise and Porcupine Abyssal Plain showed anomalously high U contents in sediments laid down during the last glacial stage (radiocarbon age 12-24 kyr B. P.; Thomson et al., 1990). AH the cores exhibit maximum Mn levels and maximum U levels in the same sequence with increasing depth in cores. However, no correlation is observed between authigenic U and organic carbon contents of these cores. Thomson and his colleagues (1990) suggested that the U enrichment is not syngenetic with the depositian of its hast glacial age sediments but is, rather, an early diagenetic addition. The position of the U peak is a result of the glacial stagelHolocene decrease in sediment accumulation rate and subsequent deepening of the oxic/post-oxic boundary into the glacial age sediments during Holocene pelagie deposition (Thomson et al., 1990). In the MD962085 core, no difference in terms of mean sedimentation rates between the last glacial stage and the Holocene was found. Thus, as already discussed in the previous section of this paper, the higher paleoproductivity due ta more intensive upwelling of the Benguela upwelling system may be the major cause of high authigenic U concentrations during the last full glacial stage in this region.
Although glacial high levels of authigenic U in the MD962085 core are largely caused by high export flux of organic material, the oxygen concentration in the water columns may play an additional role in enhancing this high level phenomenon. The 02-depleted gyre water of the Angola Current (Visser, 1969), and more intensive upwelling activity of the Benguela upwelling system during the glacial periods may have caused the reduced condition to develop not only at the bottom water-sediment interface, but also in the water column and in microenvironments in the settling partic1es. Ta prove whether this alternative process is an important mechanism for the removal of uranium from sea water into the sediments, an experiment examining the uranium content of particulate matter collected with sediment traps is needed.