RESULTS AND DISCUSSION

All inorganic geochemical data discussed in this publication are listed in Table T1. The most important aspects of the data set are presented and discussed below.

The investigated sediments may be regarded as clay or calcareous nannofossil-rich clay with carbonate contents between 5 and 30 wt%. According to our own data (Table T1) as well as shipboard measurements (Shipboard Scientific Party, 2000), the organic carbon contents vary between 0.2 and 0.5 wt%, suggesting permanently oxygenated conditions at the seafloor and a moderate organic carbon flux.

In a ternary plot (Fig. F3), which shows the relative proportions of SiO₂ (representing quartz or opaline silica), Al₂O₃ (representing clay minerals), and CaO (representing carbonate), the sediments can be described as mixtures of biogenous carbonate with an aluminosilicate component that has a slightly higher Al content than average shale. Sediments and soils consisting of highly weathered material—predominantly found in tropical regions—are known for their high Al contents (Mason and Moore, 1985). Because there are no samples plotting toward higher SiO₂ contents in the ternary plot, quartz and biogenic silica fractions are rather low or negligible (Fig. F3).

In contrast to sediments from ODP Site 1145, which show a certain eolian (loess) influence (Wehausen and Brumsack, 2002), sediments from Site 1143 presented here are characterized by relatively low TiO₂/Al₂O₃ ratios (Fig. F4). Whereas Site 1145 samples plot closer to the dilution line for loess (Schnetger, 1992), Site 1143 samples plot toward higher Al₂O₃ values, relatively close to the data point for Mekong suspended matter (SPM) (Martin and Meybeck, 1979). However, in addition to material from the Mekong River, there seems to have been at least one other source with slightly higher Ti and lower Al contents delivering material to this part of the South China Sea because all samples plot above the ratio for Mekong SPM (Fig. F4).

Depth profiles of CaCO₃, SiO₂, and Al₂O₃ contents demonstrate that the bulk composition of the sediments show a high variability with time (Fig. F5). SiO₂ and Al₂O₃ display the same pattern (correlation coefficient: r² = 0.92), which again shows that there is no significant contribution of biogenic opal to the total SiO₂ content. Only very few diatoms and radiolarians were found in the late Pliocene sediments of Site 1143 (Shipboard Scientific Party, 2000).

With regard to the carbonate contents, two different types of carbonate cycles have been reported for South China Sea sediments (Thunell et al., 1992). Carbonate cycles with higher carbonate contents during glacials ("Pacific type"), reflecting the varying carbonate compensation depth (CCD) or lysocline, are only observed in sediments deposited in water depths below 3500 m. Classical carbonate cycles, showing lower carbonate content during glacials due to dilution by terrigenous detrital material and higher carbonate contents during interglacials ("Atlantic type") (Thunell et al., 1992), are found in water depths above 3000 m. At present, the water depth of Site 1143 is 2772 m. Assuming a similar depth during the Pliocene (permanent sedimentation above the CCD), "Atlantic-type" carbonate cyclicity should be expected at this location. Our data show that carbonate contents are generally lower and terrigenous detrital matter (SiO₂ and Al₂O₃) contents are higher during glacial stages (Fig. F5). This supports "Atlantic-type" carbonate cycles mode (i.e., dilution cycles). Minimum carbonate and maximum terrigenous detrital matter contents are observed for oxygen isotope Stages 100, 104, 106, and 110. Beyond some small peaks in sediment composition (i.e., changes of short duration), the carbonate terrigenous detrital matter relationship displays low variability before 2.85 Ma (Fig. F5).

In order to define changes in the composition of the terrigenous detrital matter, we either use elemental contents that were calculated on a carbonate-free basis (see "Materials and Methods") or ratios of terrigenous detrital elements.

In general, slight changes in the composition of terrigenous detrital matter are discernible. A few distinct thin layers show significantly lower contents of most major components of the noncarbonate fraction (Fig. F6). TiO₂ (cfb) displays a relatively strong variation throughout the investigated interval. A shift toward lower TiO₂ (cfb) values occurs at ~2.9 Ma. SiO₂ (cfb) and K₂O (cfb) covary more or less with TiO₂ (cfb) but show relatively little variability. The Al₂O₃ (cfb) content displays only small variations in sediments older than 2.6 Ma. In the upper part of the core, variations are stronger with an opposite pattern compared to SiO₂ (cfb), TiO₂ (cfb), and K₂O (cfb) (Fig. F6A).

The Si/Al ratio shows low variations in sediments >2.6 Ma and higher variation in sediments <2.6 Ma (Fig. F6A). In contrast, Ti/Al displays strong variations throughout the whole section (Fig. F6B), mainly driven by the TiO₂ (cfb) content with relatively high amplitude variations (Fig. F6A). K/Al, Cr/Al, Rb/Al, and Zr/Al ratios covary more or less with the Ti/Al ratio, and all five parameters display long-term changes. They decrease between 2.9 and 2.5 Ma and slightly increase again thereafter.

To observe and interpret those changes in terrigenous detrital matter composition with respect to global ice volume and sea level changes, we compared the Al₂O₃ (cfb) and TiO₂ (cfb) records with the benthic ¹⁸O curve (Fig. F7). Higher contents of Al₂O₃ (cfb) are present during glacial stages owing to an enhanced contribution of strongly weathered material from the Asian continent or adjacent islands. TiO₂ (cfb) contents, which are lower during glacials, are due to stronger fluvial input of material with low Ti content (e.g., suspended matter from Mekong River) (Martin and Meybeck, 1979). Until 2.8 Ma, a significant increase in Al₂O₃ (cfb) content is not seen. A weak enrichment in Al is visible at oxygen isotope Stage 110 (2.73 Ma), which marks the onset of major glaciation cycles of the Northern Hemisphere (Tiedemann et al., 1994). Amplitudes of Al₂O₃ (cfb) content increase after 2.55 Ma, and the strongest Al enrichments are present during isotope Stages 100, 96, and 82. The highest amplitudes and, thus, the strongest minima are present in the TiO₂ (cfb) record between 2.45 and 2.8 Ma.

What are the climatic and/or oceanographic mechanisms causing changes in terrigenous detrital matter composition? There are four possible scenarios, as known from earlier publications and described in the following paragraphs.

1. Changing weathering conditions on the continent. Sediments from the Zaire Fan display cyclic changes in the Al/K ratio, which Schneider et al. (1997) interpreted as fluctuations in weathering intensity related to variations in the West African monsoon. The Al/K ratio served as an indicator for the dominance of kaolinite, an important product of intensive weathering, over feldspar (or other K-bearing minerals). By applying this geochemical indicator, it was demonstrated that increased central African heating during astronomical insolation maxima led to a stronger monsoonal precipitation (Schneider et al., 1997).
    

   The chemical index of alteration (CIA) (Nesbitt and Young, 1982) in sediments from the Bay of Bengal was found to be related to chemical weathering driven by the Asian summer monsoon (Colin et al., 1998). The CIA is calculated from Al, Ca, Na, and K contents.
    

   These two examples show that monsoon-related weathering intensity can affect the relative abundances of major elements in the terrigenous detrital fraction of sediments. For the South China Sea, weathering may also explain changes in the Al₂O₃ (cfb) contents as seen in Figure F7 but will probably fail to explain, for example, SiO₂ (cfb) and TiO₂ (cfb) profiles. These show distinct patterns that can not be explained by variations in weathering-related clay mineral composition (i.e., Al, Na, or K contribution) alone. If weathering would be the only mechanism causing differences in the composition of the terrigenous detrital matter composition, then SiO₂ (cfb) and TiO₂ (cfb) profiles would probably display an exactly reverse pattern compared to the Al₂O₃ (cfb) profile because of the closed sum effect within the terrigenous detrital matter fraction. Owing to their occurrence in more weathering-resistant minerals (e.g., quartz, rutile, and ilmenite), Si and Ti contents are not dominated by weathering effects. Therefore, these two elements should not display a distinct pattern, as is the case for the sediments investigated here. As a consequence, changes in weathering intensity may only form one part of the answer.

2. Variations in provenance (e.g., stronger riverine contribution during maxima in summer monsoon). Several studies demonstrated that differences in provenance are seen in the major element geochemistry of sediments. For example, Shimmield and Mowbray (1991) were able to identify monsoon-related changes in the eolian flux into Arabian Sea sediments through elemental ratios like Ti/Al and Cr/Al. Similarly, Wehausen and Brumsack (1999) used Mg/Al, K/Al, and other element ratios to show that the relative contribution of eolian and fluvial sources to the eastern Mediterranean varied due to more arid or humid climate conditions on the African and southern European continents. For the Site 1143 data discussed here, fluctuations in the TiO₂ (cfb) content, K₂O (cfb) content, Cr/Al ratio, and Rb/Al ratio may also be explained by two different sources (e.g., either the Mekong River or rivers from Borneo delivering more or less material depending on monsoonal variations). However, if there were in fact two (or more) different sources, they should be characterized by material that is very similar in composition because the detected variability in sediment composition is only of minor importance.
    

   In contrast to Arabian Sea (Shimmield and Mowbray, 1991) or eastern Mediterranean (Wehausen and Brumsack, 1999) sediments, Site 1143 sediments are strongly dominated by fluvial input, because the eolian contribution is low for this part of the South China Sea (see "Sample Description" in "Materials and Methods"). Compared to eolian fluxes (Rea, 1994), fluvial contributions to distal sites like Site 1143 depend much more on water currents as well as basin and river fan structures, both of which are influenced by sea level. Therefore, source strength cannot be assessed adequately without discussing the following two points.

3. Changes in the strength and direction of water currents. The influence of surface-water circulation on sediment dispersal has been discussed in earlier publications. One example seems to be reflected in sediments from the Ceara Rise, in the equatorial western Atlantic (Tiedemann and Franz, 1997). Here, the current system off Brazil, which is mainly influenced by the latitudinal position of the intertropical convergence zone, has a strong influence on the transport of suspended sediment from the Amazon River to the Ceara Rise. In the modern eastern Mediterranean, the complicated current and water circulation system, for example, is responsible for a distinct dispersal pattern of terrigenous matter from several fluvial sources (Venkatarathnam and Ryan, 1971). In the past, changes in current directions seem to have led to significant changes in sediment distribution patterns (Wehausen and Brumsack, 2000).
    

   To our knowledge, data on dispersal tracks of the major riverine sediment sources of the South China Sea as related to current directions are not available. However, the surface-water circulation of the South China Sea is mainly driven by the annually reversing monsoon winds (Huang et al., 1994). During the summer monsoon, a rather clockwise surface circulation, which also reaches the bottom area of the shelf (Huang et al., 1994), may deflect the strengthened monsoonal runoff from the Mekong River toward the north. During winter monsoon, when Mekong River runoff is at its minimum, a counterclockwise circulation may deliver suspended matter from the shelf area of the southwestern part of the South China Sea to Site 1143. However, it remains questionable which of the two scenarios could have led to a higher contribution of suspended matter to Site 1143. Furthermore, the system becomes more complicated when eustatic sea level fluctuations are taken into account.
    

   Strong bottom currents may have an additional influence on sediment composition via winnowing or focusing. Such phenomena led to the loss or accumulation of the fine fraction at the sediment surface. Usually, such deposits are characterized by layers significantly enriched or depleted in heavy mineral-related elements like Zr. Contents of this element strongly depend on grain size sorting effects (Wehausen, 1999). Although turbidites are a common feature, especially in the older sediments of Site 1143 (Shipboard Scientific Party, 2000), no such layers were detected in the core sections investigated in the present work.

4. Sea level changes. Because of the presence of a large shelf area in its southern part, the South China Sea became a semienclosed basin during the last glacial maximum with significant consequences regarding water circulation patterns (Wang and Wang, 1990). At that time the drastic sea level lowering also led to erosion of material from the exposed shelf. This material was transported via the Paleo-North Sunda River system (Wang et al., 1995) more or less directly into the present deep central basin. As another consequence, material from the exposed Sunda shelf, with a different composition compared to material from Mekong River, presumably has influenced terrigenous detrital matter composition at Site 1143. Additionally, the Mekong River delta shifted eastward, probably also leading to a higher suspended matter contribution to Site 1143. The overall consequence must have been an increase in terrigenous detrital matter accumulation in the area around Site 1143.
    

   Although sea level changes as related to global ice volume were much weaker during the late Pliocene (Tiedemann et al., 1994), certain fluctuations in the size of the shelf area and the position of the Mekong River delta must have occurred. During glacial stages, a slightly enhanced sediment discharge is seen. This explains the higher contents of terrigenous detrital matter and lower contents of carbonate (Fig. F5). Perhaps it may also explain the higher Al₂O₃ (cfb) contents because Mekong riverine suspended matter is relatively enriched in Al when compared to crustal or average shale abundances (Taylor and McLennan, 1985; Wedepohl, 1971).
    

   However, this scenario alone does not account for differences in detrital element profile patterns and global ice volume. At least a combined influence of two different mechanisms, either climate related (changes in weathering intensity or varying sources) or oceanography-related (changes in sea level, basin structure, or water currents), must have operated simultaneously.

The Ba content calculated on a carbonate-free basis or the Ba/Al ratio are parameters that may serve as indicators for paleoproductivity (Dymond et al., 1992; Francois et al., 1995). Both show exactly the same cyclic profile, but we prefer to use Ba/Al ratios for the sake of a better comparability with other studies where this proxy has been applied (Wehausen and Brumsack, 1999; Shimmield and Mowbray, 1991). Before using such Al-normalized total Ba contents as a proxy for marine productivity conditions, it has to be evaluated whether the fluctuations are solely related to changes in terrigenous input, are only caused by variations in barite preservation (Schenau et al., 2001), or are indicative of real changes in barium flux rates and oceanic productivity. From other marine geochemical studies (Wehausen and Brumsack, 1999; Shimmield and Mowbray, 1991), as well as from the composition of the Earth's upper crust (Taylor and MacLennan, 1985) or average shale (Wedepohl, 1971, 1991), we know that the background Ba/Al value, at maximum, should be 70 x 10^-4. Because of the relatively high Al contents typical of a region with intense tropical weathering (see "Changes in Terrigenous Detrital Matter Composition"), the background value for the South China Sea is lower than it is for average shale or sediments from higher latitudes. The Ba/Al variability of the investigated sediments is higher than any reasonable lithogenic background value. Therefore, the peaklike Ba enrichments in the data set (Fig. F8) are likely caused by an enhanced bio-barite flux (Dehairs et al., 1980; Bishop, 1988; Gingele and Dahmke, 1994) and thus could be used as a proxy for paleoproductivity. Certainly, the preservation of barite as determined by the degree of saturation in bottom and pore waters is an important factor influencing the Ba contents of marine sediments (McManus et al., 1998; Schenau et al., 2001). However, when pore water sulfate concentrations are more or less constant (i.e., under permanently oxic conditions), the degree of saturation itself is only controlled by the flux of barite (which acts as a positive feedback). Therefore, we do not believe that productivity and the associated barium flux were only constant or even lower when Ba enrichments are present in the sediment. Instead, a positive relationship between productivity and barium content of the sediments is most likely.

The Ba/Al ratio corresponds very well to the ¹⁸O curve. This indicates that productivity was lower during glacial stages and higher during interglacials. Exceptions are seen in the lower part of the core (during stages 110 and 120, productivity was slightly higher). The covariation between the Ba/Al ratio and carbonate content suggests that increases in carbonate contents during interglacial stages are not only due to less dilution but are also caused by enhanced biological production. Shimmield and Mowbray (1991) found a very good correlation between ¹⁸O of planktonic foraminifers, Ba/Al, and carbonate records for late Quaternary sediments from the northwest Arabian Sea. They proposed that the nutrient supply through upwelling of intermediate waters was the triggering factor for higher productivity during interglacials. A similar explanation may be valid for Site 1143 sediments, although we cannot solve the paleoclimatic and paleoceanographic mechanism causing a higher nutrient supply during interglacials in the southern South China Sea. One possibility may be the stronger inflow of nutrient-rich waters from the Sunda shelf. Another possibility is enhanced upwelling caused by stronger monsoonal winds.

The P/Al ratio displays a cyclic record that correlates with Ba/Al and carbonate contents (Fig. F8). The primary flux of phosphorus into marine sediments mainly appears in three different forms: organic material, fish remnants, and iron oxides that have a high adsorption capacity for phosphorus (Froehlich et al., 1988; Van Cappellen and Berner, 1988; Van Cappellen and Ingall, 1994). Lithogenic phosphorus is generally of minor importance. In deeper buried sediment layers phosphorus may still be present in these forms, but here, diagenetically formed carbonate-fluoroapatite (CFA) is responsible for the major fraction of phosphorus (Ruttenberg and Berner, 1993). About 80% of the total P in Pliocene sediments from various settings were found to be present as CFA (Delaney, 1998).

Despite the dominance of authigenic phosphorus phases in deeper buried sediments, the covariation of P/Al ratio, carbonate content, and Ba/Al ratio suggests that the primary availability (i.e., changes in bioproductivity and phosphorus flux) had a significant influence on the phosphorus content of Site 1143 sediments. Another important factor for phosphorus burial is the carbonate content because the high surface area and adsorption capacity of calcite might trigger the formation of iron oxyhydroxide coatings, adsorbing agents for phosphorus (Delaney, 1998).

In general, Fe₂O₃ (cfb) contents display variations that covary with those of SiO₂ (cfb) (Figs. F9A, F6A). Besides these minor fluctuations caused by changes in terrigenous detrital matter composition, there are some small spikes and three very prominent peaks. These layers are characterized by lower terrigenous detrital matter contents (Fig. F6A) and enrichments in iron, sulfur, and the trace elements As, Co, and Ni (Fig. F9B). These trace elements are known to form stable sulfides or to coprecipitate with iron sulfides under sulfidic conditions (Jacobs and Emerson, 1985; Huerta-Diaz and Morse, 1992; Calvert and Pedersen, 1993). Vanadium, mainly present in the form of vanadyl cations in organic complexes under reducing conditions (Szalay and Szilagyi, 1967; Emerson and Huested, 1991), does not show a significant enrichment. All of this suggests the presence of pyrite and other metal sulfides of diagenetic origin in the form of small concretions that seem to have formed very locally. Evidence for microbial sulfate reduction was given by the pore water data (Shipboard Scientific Party, 2000).

High Mn/Al ratios well above the ratio for average shale and a partial covariation between Mn and carbonate contents (Fig. F9A) suggests the presence of manganese coatings on carbonate tests (Boyle, 1983; Franklin and Morse, 1983) or the presence of authigenic manganese carbonates (Thomson et al., 1986). Whereas Sr clearly correlates with CaCO₃ (r² = 0.96), Mn/Al displays a more specific behavior, probably caused by changes in terrigenous detrital matter flux. Furthermore, some enrichments in manganese are seen that cannot be explained by carbonate contents or terrigenous input. The Mn/Al peaks at 2180, 2210, 2260, and 2430 ka may, for example, represent periods of a strong import of Mn from the shelf. Such an Mn accumulation mechanism has been reported for the Cretaceous Indian Ocean (Thurow et al., 1992). Oxygen-deficient conditions led to the mobilization of Mn(II) from shelf sediments and transport via the oxygen minimum zone into deeper regions of the basin. Since the Mn enrichments occurred during interglacials and corresponding times of high bioproductivity, the existence of an enhanced oxygen minimum zone is very likely.