The radiogenic isotope composition of dissolved trace metals in the ocean represents a set of relatively new and not yet fully exploited tracers with a large potential for oceanographic and paleoceanographic research on timescales from the present back to at least 60 Ma. The main topic of this review are those trace metals with oceanic residence times on the order of or shorter than the global mixing time of the ocean (Nd, Pb, Hf, and, in addition, Be). Their isotopic composition in the ocean has varied as a function of changes in paleocirculation, source provenances, style and intensity of weathering on the continents, as well as orogenic processes. The relative importance of these processes for each trace metal is evaluated, which is a prerequisite for reliable interpretation of their time series in terms of changes in paleocirculation or weathering inputs. This analysis of processes includes a discussion of the long-term isotopic evolution of Sr and Os, which are well mixed in the ocean and have thus not been influenced by circulation changes. The radiogenic isotope evolution of those trace metals with intermediate oceanic residence times can be used as paleoceanographic proxies to reconstruct paleocirculation and weathering inputs into the ocean. This is demonstrated by studies from different ocean basins, mainly carried out on ferromanganese crusts, which show that radiogenic trace metal isotopes provide important new insights and can complement results obtained by other well-established paleoceanographic tracers such as carbon isotopes.
 In a simplified model the modern global ocean circulation system is largely driven by the sinking of cold, saline (and therefore dense) water masses at high geographical latitudes [cf. Gordon, 1986; Broecker, 1991; Schmitz, 1995]. The most important location for deep-water production is the North Atlantic Ocean, where essentially four different sources (Iceland Scotland Ridge Overflow Water and Denmark Strait Overflow Water, which both originate from downwelling in the Norwegian/Greenland Seas, Labrador Seawater, and Lower Deep Water, which is derived from the Southern Ocean) produce a flow of some 15±2 sverdrup (Sv) (1 Sv = 106 m3−1) of North Atlantic Deep Water (NADW) according to latest estimates based on results of the World Ocean Circulation Experiment [Ganachaud and Wunsch, 2000] (Figure 1). NADW has a characteristic temperature and salinity and flows south, thereby entraining northward flowing Antarctic Bottom Water (AABW) from below and Antarctic Intermediate Water (AAIW) from above resulting in an increased flow of 23±3 Sv when it leaves the South Atlantic at 30°S and joins the Antarctic Circumpolar Current (ACC) system and its main water mass Circumpolar Deep Water (CDW). The ACC flows eastward at a rate of 140–150 Sv, which makes it the largest current of the ocean in terms of volume transport.
 The Antarctic ice shelves represent a second important source of deep water. There, deep water is produced by formation of sea ice, which consists of freshwater and leaves a highly saline and cold brine behind, which sinks because of its high density. Through mixing of this water with other water masses, mainly CDW, 21±6 Sv AABW are formed, which is denser than NADW. Bottom water originating from the Southern Ocean (eventually a mixture of AABW and NADW) flows northward in all three major ocean basins where the net inflow amounts to 6±1.3 Sv into the Atlantic Ocean, 11±4 Sv into the Indian Ocean, and 7±2 Sv into the Pacific Ocean [Ganachaud and Wunsch, 2000]. The northward flow of these deep and bottom water currents into the Pacific and Indian Oceans is mainly balanced by southward flowing intermediate and deep waters. Major upper level currents balancing mass flux at depth are the Pacific-Indonesian throughflow at a rate of 16±5 Sv and thermocline waters from the Southern Ocean, which flow north in the Atlantic Ocean at a rate of 16±3 Sv (AAIW and eventually the Gulf Stream surface current). Together these currents, whose detailed distribution and water mass exchange is much more complicated in detail, form the global thermohaline circulation. This system largely controls the transfer of heat and moisture on Earth and is thus one of the most important factors controlling global climate. There is evidence that the thermohaline circulation system is more complex than previously thought in that there may not only be one single global thermohaline circulation cell, but that two nearly independent cells coexist in which turbulent flow plays an important role [Macdonald and Wunsch, 1996]. In addition, it has recently been speculated that diapycnal mixing rates (vertical diffusive mixing across water mass boundaries) may have been overestimated in ocean circulation models, which is important with respect to air-sea gas exchange [Archer et al., 2000]. Nevertheless, the overall general circulation patterns and amounts of water mass exchanged in the present-day ocean can be considered valid and will serve as a basis for the interpretation of the radiogenic isotope results.
1.2. Ocean Circulation in the Past
 The current mode of ocean circulation has gone through substantial changes in the past. On decadal timescales, for example, a strongly reduced deep-water production in the Greenland Sea was observed during the 1980s which might have been caused by anthropogenic perturbations of the climate system [Schlosser et al., 1991]. For reconstructions of the distributions and flow patterns of water masses in the more distant past, classic tools of oceanographers, temperature and salinity, which are conservative in the ocean, as well as nutrient concentrations, such as phosphate, nitrate or silica, which are typical of certain water masses, are not preserved. Therefore a number of proxy tracers have been developed to extract information on water mass distribution and flow in the past from the marine sediments. The most commonly applied geochemical proxies to reconstruct past deep-water circulation have been the stable carbon isotope composition (δ13C) given in per mill:
[cf. Curry and Lohmann, 1983; Curry et al., 1988; Woodruff and Savin, 1989; Raymo et al., 1992; Sarnthein et al., 1994; Ravelo and Andreasen, 2000] and the Cd/Ca ratio [Boyle, 1988, 1992] as recorded and preserved by benthic foraminifera. The δ13C of deep water is generally anticorrelated with nutrient (phosphate) concentration because of the preferential incorporation of 12C into organic matter (thus low in δ13C) at the surface of the ocean which is released together with nutrients during organic matter remineralization at depth. Cd/Ca correlates positively with nutrient concentrations (phosphate, nitrate). This allows the distinction between low-nutrient (such as NADW) and high-nutrient deep waters (such as in the North Pacific) in the past. For example, comparison of the δ13C gradients between the Atlantic and Pacific revealed a major reduction in the vigor of the global thermohaline during the glacial periods of the Pleistocene [Raymo et al., 1990]. The same result was also obtained from reconstructions of the oceanic distribution of 14C during the last glacial period [Broecker et al., 1990].
 It is, however, difficult to make quantitative estimates of mixing between water masses on the basis of Cd/Ca or δ13C data because neither of the two proxies may exclusively mirror the nutrient content of ambient deep water. In the case of δ13C, nonconservative effects of temperature and nutrient availability (see overview by Broecker and Peng ) or variability in carbonate ion concentration [Spero et al., 1997] are potentially superimposed. For Cd/Ca, problems arise from thermodynamic effects [Boyle, 1988]. Nevertheless, δ13C and Cd/Ca have clearly been the most reliable proxies for past water mass distribution, which have been improved and refined over the past two decades. It is those tracers from which most of our present knowledge on past ocean circulation has been derived, and in most areas of the world ocean they yield consistent results. There are only few areas where results obtained from δ13C and Cd/Ca are not consistent, which probably reflects some of the effects described above. One example is the Southern Ocean, where Cd/Ca shows no differences between the glacial and interglacial deep-water signal [Boyle, 1992], but δ13C indicates a major decrease or even shut-down of glacial NADW input into the Southern Ocean [Charles and Fairbanks, 1992]. In addition, Cd/Ca and δ13C are tracers of nutrient concentrations, which are controlled by processes in the surface ocean. Thus deep-water masses of different origins within an ocean basin may have essentially the same Cd/Ca and δ13C signatures but cannot be distinguished further without additional information.
 An alternative approach to evaluate the strength of the global thermohaline circulation on glacial-interglacial timescales has involved the use of the 231Pa/230Th ratio adsorbed to marine sediment particles [Yu et al., 1996]. From similar glacial and interglacial amounts of 231Pa exported from the Atlantic basins into the Southern Ocean, Yu et al.  concluded that there was no significant difference between glacial and interglacial strength of NADW production and thus thermohaline circulation. However, it was later suggested that the 231Pa/230Th ratio in the Southern Ocean may be less sensitive to changes in NADW export than previously suggested, thus preventing a quantitative reconstruction of water mass mixing with this proxy [Asmus et al., 1999]. In addition, it has been claimed that the uncertainties of the mean Atlantic 231Pa/230Th ratios of the Holocene and the Last Glacial Maximum are too large to rule out even large changes in NADW export [Marchal et al., 2000].
 On Cenozoic timescales, paleogeographic changes, such as openings and closings of oceanic gateways (Figure 2), have been responsible for large-scale reorganizations of the global ocean circulation and were intimately linked to global climate (see Zachos et al.  for a recent summary). Tectonically induced initiation (cessation) of water mass exchange between ocean basins caused converging (diverging) evolution of organisms in the respective basins and was also recorded by changes of geochemical parameters such as the corrosiveness of deep waters to carbonate. One of the most important openings that led to dramatic changes of the global ocean circulation in the course of the Tertiary was the stepwise establishment of the ACC. This started by rifting and subsequent northward drift of Australia away from the Antarctic at about 55 Ma and led to the establishment of a deep-water passage, the Tasman Strait, at about 30–25 Ma [Weissel and Hayes, 1972; Kennett, 1977]. The opening of Drake Passage between South America and the Antarctic at roughly 23 Ma [Barker and Burrell, 1977] completed this process, which eventually permitted unrestricted water mass exchange around the Antarctic, thermal isolation, and, ultimately, continuous Antarctic glaciation since about 14 Ma [cf. Kennett, 1977]. Over the period between 25 and 5 Ma other low-latitude gateways such as the Tethyan passages and the Indonesian seaway were closing with respect to deep-water circulation (Figure 2). The most recent major event was the progressive shallowing and closing of the Panama gateway, which was finalized by the emergence of the Isthmus of Panama, at about 3.5 Ma [Keigwin, 1982]. This stopped the direct exchange of Pacific and Atlantic water masses at low latitudes.
 Mechanisms of deep-water production prior to these paleogeographic changes (for example during the Eocene) were completely different than today. The global circulation system was dominated by meridional currents, and no significant ice caps at the poles existed. The ocean was ventilated by halothermal processes, whereby warm and highly saline oxygen-rich water masses sank at low latitudes, as evidenced by estimates of deep-water temperatures of up to 12°–14°C during the Eocene period at about 50 Ma [Miller et al., 1987; Lear et al., 2000]. The following paleogeographic reorganizations during the Tertiary caused a stepwise transition from a global ocean circulation system dominated by meridional currents at low latitudes toward the more latitudinally dominated circulation system of the present-day with its production of cold deep waters at high latitudes [cf. Haq, 1981].
 Related to these circulation changes, global climate has shown a general cooling since the Eocene climate optimum at circa 50 Ma. This general trend has been punctuated by steps during which the global climate exceeded thresholds permitting the first Antarctic glaciation at the Eocene-Oligocene boundary, the establishment of the present Asian monsoonal system at about 7 Ma, and the initiation of Northern Hemisphere glaciation and its glacial-interglacial cyclicity at 3 Ma [cf. Zachos et al., 2001]. These steps were also related to the uplift of major orogens such as the Himalayas, with particular importance of the Tibetan Plateau, and the Andes, which had major consequences for the global atmospheric circulation patterns, the distribution of the climate zones, and the intensity and style of weathering on the continents [e.g., Raymo and Ruddiman, 1992].
1.3. Radiogenic Isotopes as Water Mass Proxies
 The radiogenic isotope ratios of trace metals dissolved in seawater discussed in this review are independent of fractionation induced by biological processes or evaporation. As a consequence, a change in the trace metal isotope composition of a water mass can only occur by addition from a reservoir of the respective trace metal with a different isotope composition. The factors other than mixing controlling the radiogenic isotope composition of a water mass are completely independent from the processes controlling δ13C or Cd/Ca, and thus radiogenic isotopes can provide useful additional tracer information on past water mass exchange. The focus of this review will be on trace metals with an ocean residence time on the order of or shorter than the average circulation time of the global ocean, approximately 1500 years. If the residence time of a metal in the ocean is very short, for example, that of Th is only 5–20 years, it will not be suitable for this approach because water mass bodies will not be able to develop a typical signature which is traceable over longer distances. If, on the other hand, the residence time is too long, as is the case for U, Sr, and Os with oceanic residence times on the order of tens to hundreds of kiloyears or even million years, any water mass signatures will have been removed by mixing. In addition, the half-lives of any radioactive trace metal isotopes (10Be) have to be long relative to the mixing times, at least within an ocean basin. The half-life obviously also has to be long enough to allow for the variability in trace metal isotope composition to be preserved on the timescales of the paleoceanographic reconstruction. These requirements leave the following metals suitable for this study: Nd, Pb, Hf (for these three metals all applied isotopes are stable), and the ratio between radioactive 10Be and stable 9Be (Table 1). For a reliable reconstruction of paleoceanographic changes using radiogenic isotopes, the evolution of the long residence time radiogenic isotope systems of Os and Sr provide valuable information on changes of the global weathering inputs, which have potentially also influenced the signals of the shorter residence time tracers. Os and Sr isotopes will therefore also be discussed.
Table 1. Radiogenic Isotope Systems Used as Tracers in the Ocean
The primordial isotopes printed in boldface are those commonly used for the radiogenic isotope ratios.
The isotope 186Os has previously been used for the isotope ratios as well.
The isotope 10Be is not formed by decay of a parent isotope but is radioactive itself. Nevertheless, the ratio between 10Be and 9Be in the ocean can be used in a similar way as the radiogenic isotope systems for the past 10 Myr.
1.4. Origin of Radiogenic Isotope Variability in the Ocean
 During extraction of the continental crust from the mantle and subsequent reworking during Earth's history, elements have been fractionated. Some of these elements are radioactive and produce daughter isotopes of other elements. As an example, Sm/Nd ratios in Earth's mantle and thus also mantle-derived rocks such as mid-ocean ridge basalts are generally higher than in the continental crust because Sm stays preferentially in the mantle during the formation of continental crust. 147Sm is radioactive and produces stable 143Nd at a very slow rate (the half-life of 147Sm is 106 Gyr). Consequently, the abundance of 143Nd in a rock relative to those of other stable isotopes of Nd, such as 144Nd, vary as a function of age and Sm/Nd ratio.
 Similar to Nd, there is a stable Hf isotope, 176Hf, which causes variations in the 176Hf/177Hf ratio (177Hf is primordial) between different rocks because of radiogenic ingrowth from decay of its parent isotope 176Lu (half-life is 37.3 Gyr). For Pb, there are three stable isotopes (206Pb, 207Pb, 208Pb) which are the final products of the three U decay series with the parent isotopes 238U, 235U and 232Th (half-lives are 4.47 Gyr, 0.704 Gyr, and 14 Gyr, respectively). In addition, there is primordial 204Pb. Os shows variations in its 187Os/188Os ratio (188Os is primordial) between rocks due to decay of the parent isotope of 187Os, 187Re (half-life is 43 Gyr). Analogous to Nd, the isotope ratios of Pb, Hf, Os, and Sr in crustal rocks vary as a function of age and the ratios between daughter and parent isotope abundances.
 The release of trace metals with different isotope ratios from Earth's crust by weathering or hydrothermal processes is the basis for the observed variability in the ocean's water masses. Because of the very small differences among 143Nd/144Nd ratios and 176Hf/177Hf ratios, which are mostly at the fourth or even fifth decimal place, Nd isotope ratios are expressed as ϵNd values:
where CHUR is the 143Nd/144Nd of the chondritic uniform reservoir (presently 0.512638). Analogous to Nd, Hf isotope ratios are given as ϵHf values, with a 176Hf/177Hf for CHUR of 0.282769 [Nowell et al., 1998]. CHUR stands for a model describing the terrestrial Nd and Hf isotope evolution from a uniform reservoir whose Sm/Nd and Lu/Hf ratios are equal to those of chondritic meteorites. All time series of ϵNd and ϵHf in this review have been corrected to the corresponding value of CHUR at the time of their deposition in the marine archives.
 The variability of the Be isotope ratio, 10Be/9Be, in the ocean is mainly caused by oceanic processes (see section 3.5).
1.5. Analytical Techniques
 Following chemical separation, preconcentration, and purification, the radiogenic isotope ratios of the trace metals (on amounts as low as a few nanograms) are determined at high precision by various techniques of mass spectrometry. It is self-evident that measurement precision needs to be significantly better than the natural variability of isotope ratios. For many years, thermal ionization mass spectrometry (TIMS) has been the most commonly applied method to measure isotope ratios and metal concentrations at high precision (Nd, Sr, Pb). For some metals such as Pb the precision of the TIMS measurements has been greatly improved using double- or triple-spike methods [Galer, 1999], which is, however, very labor-intensive. Over the past few years, multiple-collector–inductively coupled plasma mass spectrometry (MC-ICPMS) has facilitated the precise measurement of isotope ratios of metals which are difficult to ionize thermally, such as Hf [ Rehkämper et al., 2001], and allows greater throughput of samples. In the case of metals that have only one or two stable or primordial isotopes, such as Cu or Pb, isotope ratios can be measured more precisely than by conventional TIMS techniques by correction for instrumental mass fractionation through addition of a spike of a different element with similar mass [Belshaw et al., 1998]. Coupled to a laser system, isotope ratios of trace metals can also be measured directly from certain materials by MC-ICPMS without chemical pretreatment [cf. Rehkämper et al., 2001]. Ratios and concentrations of other very low abundance isotopes such as 10Be are measured by accelerator mass spectrometry (AMS) [cf. Suter, 1992] or secondary ion mass spectrometry (SIMS) [Belshaw et al., 1995].
2. SOURCES OF RADIOGENIC TRACE METALS IN THE OCEAN
 There are three main pathways by which trace metals with distinct isotope compositions can be made available as dissolved tracers in the ocean (Figure 3). The first possibility is the introduction via rivers, either in dissolved, colloidal, or particulate form. The second is eolian input of particles mobilized from the continental crust by erosion or volcanic activity. Particles introduced into the ocean partially dissolve, the relative amount depending on lithology of the dust and respective trace metal [e.g., Duce et al., 1991], and release their trace metal isotope signature to the water column. Third, for some trace metals the release by hydrothermal activity on the seafloor is an important additional source. The relative importance of these sources is different for each trace metal and depends on factors such as their mobility during weathering and erosional processes and their chemical behavior in rivers and estuaries. For some dissolved trace metals the relative present-day budgets of the input sources are well known, while for others, there are considerable uncertainties.
 For Nd, there is clear evidence that hydrothermal sources do not contribute to the dissolved seawater Nd budget because of an immediate immobilization or scavenging (adsorption and subsequent removal to the sediments) of Nd by hydrothermal precipitates very close to or even within the hydrothermal sources [German et al., 1990; Halliday et al., 1992]. There is an ongoing discussion, however, about the input mechanisms of Nd, i.e., the relative importance of eolian and riverine input. From measurements of Nd isotopes and rare earth element (REE) concentrations in sediment trap material, Tachikawa et al. [1997, 1999] suggested that up to 20% of the Nd contained in dust particles is released to seawater and that eolian dust is an important contributor to the dissolved Nd budget in the Atlantic Ocean. Goldstein et al.  arrived at a similar conclusion from isotope analyses of desert-derived dust. From REE patterns along an E-W transect across the Pacific Ocean, Greaves et al.  showed that eolian inputs are important for the dissolved Nd budget of Pacific surface waters. In contrast, Jones et al.  suggested that eolian sources do not play a major role from the apparent missing imprint of the Asian dust plume on the dissolved Nd isotope composition of NW Pacific deep waters. It is quite likely, however, that very efficient mixing of Nd in Pacific thermocline waters prevents an identification of local eolian sources in the Pacific deep waters [von Blanckenburg and Igel, 1999; Igel and von Blanckenburg, 1999].
 Riverine input, particularly in regimes of strong chemical weathering, is a very important source for the dissolved Nd budget of water masses, although most of the dissolved Nd load of rivers is precipitated in the estuarine sediments [Goldstein and Jacobsen, 1988; Elderfield et al., 1990; Ingri et al., 2000]. It was shown, however, that more than 90% of the Nd, at least in high-latitude rivers, is transported to the ocean on colloids (particle size between 0.2 μm and 3000 Dalton molecular weight) [Ingri et al., 2000; Andersson et al., 2001]. Nd can also be released from resuspended sediments [Goldstein and O'Nions, 1981] or diagenesis of particles originating from rivers [Elderfield and Sholkovitz, 1987]. Sholkovitz et al.  suggested that the Equatorial Undercurrent (EUC) in the Pacific Ocean is one example of the importance of riverine input, because this current apparently derives a large amount of its dissolved REE supply from Papua New Guinea rivers. The release of trace metals such as Nd to the ocean by leaching and mobilization from continental slope sediments has also been suggested to play a role for the Nd isotope signature of the EUC [Lacan and Jeandel, 2001] and other water masses in the vicinity of the Indonesian Island Arcs [Jeandel et al., 1998].
 Nd isotopes are generally not influenced significantly by isotopic fractionation during weathering and dissolution processes of detrital material [e.g., Goldstein et al., 1984], although it was recently proposed that there may be some fractionation effects, which preferentially release Nd with unradiogenic isotope composition (low 143Nd/144Nd) owing to the more efficient dissolution of minerals with a corresponding isotope signature. Such observations have been reported for weathering of glacial tills [Öhlander et al., 2000] and during erosion and partial dissolution of rocks that supply a boreal river in northern Scandinavia [Andersson et al., 2001]. A similar conclusion was drawn from mild acid leaching experiments of Greenland river sediments [von Blanckenburg and Nägler, 2001]. It remains to be shown, however, that these fractionation effects are of quantitative importance for the Nd isotope budget of an ocean basin.
 For Pb in the ocean the importance of the sources is different (natural sources, not anthropogenic inputs). In contrast to Nd, hydrothermal input is of at least local importance for the dissolved Pb isotope signal [Barrett et al., 1987], but it is considered to be a minor contributor (<2%) for the total oceanic budget [Chen et al., 1986]. It was concluded that riverine input plays a very important role, while dust input can explain only about 10–12% of the preanthropogenic Pb budget of the ocean [Chow and Patterson, 1962]. Chow and Patterson  also estimated a natural annual global input flux of 6.3 × 107 mol yr−1 which, assuming steady state in the ocean, must equal the output (deposition in marine sediments, section 3.2). On the basis of estimates of the global flux of eolian material into the ocean and an average amount of about 8% of this material dissolving [Duce et al., 1991], a 12% contribution by dust to the dissolved Pb budget of the ocean was estimated in accordance with the former estimate [Henderson and Maier-Reimer, 2002]. Given that a significant part of the dissolved riverine Pb is scavenged in the estuaries [ Nozaki et al., 1976], it was suggested in a more recent study that dust may still be an important factor contributing to the budget of Pb and its isotope composition of seawater, particularly in ocean areas with low riverine inputs [Jones et al., 2000].
 There is an additional process that is important for the evaluation of the sources of the Pb isotope composition of the ocean. Other than described for Nd isotopes above, Pb isotopes are fractionated during weathering of continental rocks, an effect that is also well known for Sr isotopes. In the case of Pb the radioactive decay of isotopes causes radiation damage to the crystal structure of a mineral and leaves the resulting daughter isotopes more loosely bound in the minerals or even causes a mobilization of the daughter isotopes to the mineral grain boundaries. The radiogenic isotopes of Pb (206Pb, 207Pb, 208Pb), which have been produced within the rocks by decay of U and Th, are thus easier to mobilize from rocks and minerals during weathering than nonradiogenic 204Pb [Erel et al., 1994; Jones et al., 2000]. This means that the Pb isotope composition of solutions produced during continental weathering and the dissolved Pb in the ocean may not reflect the isotope composition of the source rocks on the surrounding continents, as was demonstrated for the North Atlantic Ocean [von Blanckenburg and Nägler, 2001].
 The relative importance of different sources in controlling the dissolved Hf isotope composition of the oceans is still unclear at present, mainly because Hf isotope measurements of seawater are not available yet. As soon as such measurements are available, the whole potential of the isotopic composition of this element will be accessible. It is, nevertheless, obvious from ferromanganese crust/nodule data, which reflect the deep-water Hf isotope composition, that there is a strong effect of incongruent weathering on the dissolved Hf isotope composition of seawater. This is caused by the fractionation of Hf and Lu between sand and clay-sized particles. Lu/Hf ratios and thus also 176Hf/177Hf ratios are very low in weathering-resistant zircons, which are concentrated in the sand fraction of continental and nearshore marine sediments and turbidites. This results in much lower time-integrated 176Hf/177Hf ratios in sand than in clay. The Hf in the zircons of the sand fraction is almost unavailable to weathering and is not introduced into seawater. Consequently, the seawater data, as represented by the ferromanganese crust/nodule data, are offset toward higher ϵHf for a given ϵNd compared with crustal rock data in plots of Nd versus Hf isotopes [Albarède et al., 1998; David et al., 2001] (Figure 4). In addition to this effect, it has been suggested that hydrothermal contributions may play an important role for the dissolved Hf budget of the ocean [White et al., 1986; Godfrey et al., 1997], whereas there is no information available on the importance of riverine sources because of the lack of measurements.
 The Os isotope composition of the ocean is controlled by inputs derived from continental weathering, by alteration of abyssal peridotites, and to some extent by cosmic dust [Pegram et al., 1992; Ravizza, 1993]. The average 187Os/188Os of these inputs varies between high values around 1.54 in average upper continental crust and low values around 0.126 in peridotites and cosmic dust as a consequence of fractionation between Re and Os during the formation of continental crust. Under the assumption that the ocean is at steady state with respect to Os, the unradiogenic input sources together contribute 30% to the dissolved Os in the ocean, of which ∼14% (a total of 4% of the input) originate from cosmic dust. Seventy percent of the total budget is contributed from continental (riverine) sources to reach the present 187Os/188Os of seawater (1.06) [Levasseur et al., 1999].
 As a consequence of fractionation of Rb and Sr during the formation of continental crust the Sr isotope ratios in crust and mantle differ strongly. Thus the Sr isotope budget in the ocean is mainly a balance between riverine inputs with radiogenic (high) 87Sr/86Sr ratios (average ∼ 0.7119) and mantle-derived high-temperature hydrothermal inputs at the mid-ocean ridges with unradiogenic (low) 87Sr/86Sr ratios (average ∼ 0.7035) [Palmer and Edmond, 1989]. The isotope composition of the riverine inputs can vary significantly, mainly as a function of the amount of old marine carbonates in the drainage area, which have a high Sr content and are easily dissolved. In addition, there is a small contribution (about 10% of the global riverine flux [Elderfield and Gieskes, 1982] remobilized from marine sediments via pore waters with an average isotopic composition similar to that of modern seawater (∼ 0.7084). For the contemporaneous global oceanic mass balance of Sr (total inventory = 125 × 1015 mol) it has been suggested that the most important source is the riverine flux amounting to approximately 33 × 109 mol yr−1, followed by a hydrothermal contribution on the order of 15 × 109 mol yr−1 and about 3.4 × 109 mol yr−1 from marine pore waters [Palmer and Edmond, 1989]. Earlier estimates of the hydrothermal contribution were low (2 and 4 × 109 mol yr−1) owing to low estimates of the global high-temperature hydrothermal water flux at the mid-ocean ridges [Morton and Sleep, 1985; Goldstein and Jacobsen, 1987]. These older estimates required correspondingly lower values for the average riverine Sr isotope composition (0.7095 to 0.7106) to achieve mass balance, assuming steady state of the oceans with respect to Sr concentration and isotopic composition.
3. PRESENT-DAY DISTRIBUTION OF RADIOGENIC TRACE METAL ISOTOPES IN THE OCEAN
3.1. Nd Isotopes
 The dissolved trace metals in this review differ in their oxidation states and chemical speciation in the oxygenated water column of the ocean, which determines their reactivities and oceanic residence times [Turner et al., 1981; Bruland, 1983]. Nd exists exclusively in the +3 state in oxygenated seawater similar to all other rare earth elements (except for Ce, which can be oxidized to the +4 state). Nd is most likely present as a Nd-carbonate (NdCO3+) or Nd-sulfate (NdSO4+) complex together with traces of Nd3+ [Bruland, 1983]. It is removed from the water column by particulate scavenging (adsorption), which explains its typical water column pattern of depleted surface concentrations, which increase with depth because of desorption/disaggregation processes. The global average ocean residence time of Nd in deep waters is between 600 and 2000 years [Jeandel, 1993; Jeandel et al., 1995; Tachikawa et al., 1999] and deep-water concentrations are around 4 pg g−1. On ocean basin scales where the water mass mixing times are on the order of only few hundred years, Nd can be considered quasi-conservative. Nd is deposited in marine and estuarine sediments, which constitute its most important sink.
 As a consequence, the most negative (least radiogenic, lowest 143Nd/144Nd ratios) ϵNd values are found in water masses closest to the weathering sites of old continental rocks, such as in the Baffin Bay, where ϵNd values as low as −26 have been measured in the seawater [Stordal and Wasserburg, 1986]. These waters mix with other water masses of different origin and with higher ϵNd in the North Atlantic (see section 1.1) to form NADW. This mixture results in a present-day typical value of –13.5 for NADW, whereas water masses originating in the southern Atlantic and Southern Ocean (AAIW, CDW, or AABW) have ϵNd values between –7 and –9 [Piepgras and Wasserburg, 1982,Piepgras and Wasserburg, 1987; Jeandel, 1993]. This is a consequence of mixing between Atlantic and Pacific water masses in the Southern Ocean. Figure 5 illustrates the relationship between water masses (defined by their salinities) and ϵNd on a N-S transect in the Atlantic Ocean. The tongue of southward flowing NADW with its high salinity and low ϵNd is distinct from northward flowing AAIW above and AABW below and can be identified until 49°S. The water column of the Southern Ocean is quite homogenous in ϵNd owing to efficient vertical and horizontal mixing [Piepgras and Wasserburg, 1982]. Another water mass with distinct Nd isotope composition in the Atlantic Ocean is Mediterranean Outflow Water with an ϵNd value of –9.4 [Spivack and Wasserburg, 1988]. Pacific intermediate and surface waters are much more positive in ϵNd (between 0 and –4), while deep and bottom water masses vary around lower values between –3 and –6 [Piepgras and Jacobsen, 1988; Shimizu et al., 1994]. This has been interpreted as the admixture of northward penetrating AABW. The Nd isotope composition of Indian Ocean water masses tends to be intermediate between the Atlantic and Pacific and is relatively homogenous at ϵNd values of about −7 to −8 [Bertram and Elderfield, 1993], whereas in the vicinity of the Indonesian throughflow water masses of Pacific origin are traceable by ϵNd values between −5 and −3 [Jeandel et al., 1998].
3.2. Pb Isotopes
 Dissolved Pb exists in the ocean in a +2 state mainly as Pb2+ ion or Pb carbonate complexes (PbCO3 or Pb(CO3)22−), but biologically mediated reactions can change the speciation [Bruland, 1983]. The water column profiles of Pb concentration show increased surface water values due to atmospheric (to a large extent, anthropogenic) inputs and depleted concentrations at depth [e.g., Schaule and Patterson, 1981]. Pb is a highly particle-reactive element in the ocean [Schaule and Patterson, 1981; Cochran et al., 1990], which has been shown from extensive studies of the short-lived radioactive Pb isotope 210Pb (210Pb is produced at a relatively constant rate from the decay of 226Ra in the oceans and 222Rn in the atmosphere (see Henderson and Maier-Reimer  for a compilation)). As a consequence of its particle reactivity the residence time of Pb in deep waters is only about 50 years in the Atlantic and up to 200–400 years in the Pacific with Pb concentrations around 1 pg g−1 in the deep water. Pb is thus efficiently transferred to the ocean sediments, most likely by nonreversible particulate scavenging with some release at depth mainly by particle decomposition and remineralization.
 Similar to Nd, weathering of continental crust is the main input source of preanthropogenic natural Pb into the ocean. The young mantle-derived end-member (similar to mid-ocean ridge basalt) has, for example, 206Pb/204Pb near 18.5. The continental crust end-member is difficult to define because of the effects of incongruent weathering (see section 2.2). Weathering of these rocks results in 206Pb/204Pb ratios around 19.3 for present NADW and between 18.5 and 18.8 for deep Pacific water masses. Water masses in the Indian and Southern Oceans are intermediate between those values. A similar separation between the main deep-water masses is also observed for the other radiogenic Pb isotope pairs. It must be noted that the ferromanganese crust surfaces average over several tens to hundreds of thousands of years, which includes glacial/interglacial variations. That means that the actual preanthropogenic deep-water Pb isotope composition may have had a larger range and more extreme values for particular water masses such as NADW.
3.3. Hf Isotopes
 Dissolved Hf in oxygenated seawater is present in the +4 state. It is one of the hydroxide-dominated elements which exists as hydrolysis products such as Hf(OH)5− or Hf(OH)4 [Turner et al., 1981; Bruland, 1983]. The only reliable profiles of Hf concentrations indicate a surface water depletion and an increase with depth similar to Nd, which points to some effect of particulate scavenging [Godfrey et al., 1996; McKelvey and Orians, 1998]. This suggests that, despite the lack of direct information, the major sink for Hf in the ocean is the marine sediments. The Hf residence time in seawater has been estimated from water column studies [Godfrey et al., 1996; McKelvey and Orians, 1998]. Additional constraints have been derived from studies of ferromanganese crusts, which show a range of Hf isotope variations similar to those in Nd. Given that the range in Hf isotope values in the continental crust (the ultimate source of both Nd and Hf) is about double that of Nd, a residence time of Hf on the order of 1500–2000 years is indicated [Godfrey et al., 1997; Lee et al., 1999; David et al., 2001]. Analytical difficulties and the low concentrations (0.04–0.2 pg g−1 [Godfrey et al., 1996; McKelvey and Orians, 1998]) have so far prevented the measurement of Hf isotopes in seawater. Therefore, as is the case for Pb, the distribution of Hf isotopes in the deep ocean can presently only be derived from records preserved in ferromanganese crust surfaces (see also section 5.2).
 Continental weathering and (to some extent) hydrothermal inputs control the isotopic composition of Hf in deep waters. There is a wide range in the Hf isotope composition of continental rocks which have ϵHf values as low as −30 and mantle rocks which can be as high as +25. Because of the strong incongruent weathering effects of zircons from the continental crust (see section 2.3), seawater ϵHf is higher than expected from comparison with the corresponding ϵNd data (Figure 4). The ferromanganese crust data indicate that Pacific deep waters have generally high ϵHf values between +3 and +9, while the deep North Atlantic values range between −2 and +3. Indian and Southern Ocean values are intermediate between +2 and +5.5. Analogous to the Pb isotopes, the ferromanganese crust and nodule Hf isotope data average over relatively long periods of time and the actual values of the water masses may have a somewhat wider range and more extreme values.
3.4. Os and Sr Isotopes
 Dissolved Os in the ocean is present in the +8 state and it has been suggested from thermodynamic modeling that the main chemical species are oxyanions such as H2OsO5 or H3OsO6−, although it appears that a large part of the Os is present as organic complexes [Levasseur et al., 1998]. Dissolved Os concentrations, which have only recently been measured successfully in seawater [Levasseur et al., 1998], have been shown to be homogenous at 10.86±0.07 fg g−1, although evidence has been presented since for a somewhat lower Os concentration within the oxygen minimum zone of the Pacific water column [Woodhouse et al., 1999]. A further measurement also suggested a slight deviation of the Pacific deep-water concentration from a homogenous value [Sharma et al., 2000]. The most important sinks for Os in the ocean are reducing sediments, which are found beneath the major upwelling areas.
 The average residence time of Os in the ocean is significantly longer than the circulation time of the ocean [Ravizza and Turekian, 1992; Sharma et al., 1997], but there is a discrepancy between the residence time calculated from the marine Os mass balance (>10 kyr) and that inferred from short-term glacial-interglacial Os isotope variations found in marine sediments (3–4 kyr) [Oxburgh, 1998]. This may partly arise from an underestimation of the osmium inputs into the ocean. The short-term isotopic variability observed may be related to a postglacial meltwater spike [Oxburgh, 2001]. Most recent estimates of the residence time (6.5–15 kyr) are at the lower end of the marine mass balance–based estimates [Levasseur et al., 1998; Oxburgh, 2001]. This residence time implies that any differences from the present 187Os/188Os value of 1.06 [Levasseur et al., 1998] between water masses or ocean basins must thus be very small [Burton et al., 1999b].
 Sr is a major conservative component of seawater with a residence time on the order of several million years, which is the reason why it is homogenously distributed in seawater at a concentration of 7.6 μg g−1 and the 87Sr/86Sr in the present-day ocean is uniform at 0.70918. Sr is present in the +2 state in the ocean as Sr2+ ions. The only significant sink for Sr in the ocean is the incorporation into carbonate shells and subsequent burial in marine sediments.
3.5. Be Isotopes
 Be is present in the +2 state in the ocean, mainly as a hydrolysis product such as Be(OH)+ or Be(OH)2 [Bruland, 1983]. Because of its particle reactivity, Be shows nutrient-type concentration patterns with a generally strong surface water depletion and increased values at depth caused by particulate scavenging at the surface and release at depth through desorption and particle remineralization. Its residence time in the ocean varies between about 100 years in high particle flux areas and up to 3000 years in the central Pacific gyres [Lao et al., 1992; von Blanckenburg and Igel, 1999]. The global average oceanic residence time of Be in deep waters is about 500 to 1000 years [Segl et al., 1987; Ku et al., 1990; von Blanckenburg et al., 1996a], which is similar to or somewhat shorter than that of Nd.
 Stable beryllium, 9Be, a common trace element in the continental crust, is supplied to the ocean by weathering. A radioactive beryllium isotope 10Be (T1/2 = 1.5 Myr) is produced in the upper atmosphere at an approximately constant rate by interaction of atmospheric nitrogen and oxygen with the cosmic radiation. 10Be is mainly delivered to the ocean by wet precipitation. The main factor controlling the 10Be/9Be ratio in the ocean's water column is the 10Be concentration, which increases with the “age” of the water masses owing to remineralization processes from values around 1000 atoms g−1 water in the Atlantic to about 2000 atoms g−1 in the the north Pacific [Kusakabe et al., 1987; Ku et al., 1990]. In contrast, 9Be concentrations in deep waters are quite homogenous at values around 0.25 pg g−1. The result is a distinct 10Be/9Be ratio of water masses. NADW has a typical 10Be/9Be ratio around 0.5 × 10−7, AABW and CDW have values of about 1 × 10−7, AAIW has ratios around 1.2 × 10−7, and Pacific deep waters range between 1 and 1.4 × 10−7 [Ku et al., 1990; Xu, 1994; Measures et al., 1996].
4. ISOTOPIC EVOLUTION OF THE LONG- RESIDENCE TIME TRACERS Sr AND Os IN THE OCEAN
 The isotope composition of trace metals in seawater has been reconstructed from authigenic chemical phases in the ocean. Because of their high Sr contents, (1) carbonate shells of different organisms (foraminifera, brachiopods, belemnites, conodonts) [cf. . Veizer et al., 1999]; (2) marine biogenic barite, which is produced in chemical microenvironments in the decaying organic matter of diatoms in the upper water column of the ocean [Paytan et al., 1993]; (3) fish teeth [Staudigel et al., 1985]; and (4) some phosphatic peloids [Shaw and Wasserburg, 1985] have been used for reconstructing the Sr isotope evolution of the world's ocean over the Phanerozoic (Figures 6a and 6b). For the past 60 Myr, foraminifera have been used almost exclusively [cf. Hodell et al., 1991; Veizer et al., 1999]. The evolution of the Sr isotope composition of the world ocean is characterized by an asymmetric trough-like shape over the Phanerozoic starting with 87Sr/86Sr ratios similar to the present-day value in the Cambrian [cf. Edmond, 1992]. From the Cambrian to the late Jurassic the ratios showed a general decline to ratios as low as 0.7067, which was interrupted by large, relatively short term excursions, for example, at the end of the Ordovician, the middle Devonian, and most pronounced at the Permian/Triassic boundary. At the end of the Jurassic the trend reversed and the ratios have steadily risen since, with major steepenings between about 35 and 20 Ma and over the past 2.5 Myr. The reasons for these changes have been hotly debated, but it is clear that variations in the main input sources continental erosion and hydrothermal flux must be responsible. Such changes can be accomplished in different ways. In general, increased contributions from continental weathering via river runoff tend to raise the ocean 87Sr/86Sr, whereas increased hydrothermal exchange at the mid-ocean ridges tends to lower it. This is the reason why major plate tectonic and paleogeographical changes have been invoked as a key factor in controlling the oceanic Sr isotope composition [cf. Richter et al., 1992]. During periods of major continental breakup, increased hydrothermal activity at the ridges is expected, whereas during continental collisions and orogenies an increase in erosion rate and weathering occurs. This has led to the interpretation that major increases in the oceanic 87Sr/86Sr (such as the most pronounced rise between 35 Ma and present) have been caused by major orogenies such as Himalayan uplift and related changes in weathering regime and increases in erosional inputs [cf. Richter et al., 1992]. For the Cenozoic curve it has also been suggested that contributions from the Himalayan metamorphic core complex with its extraordinarily high Sr isotope ratio and high Sr concentrations in the rivers have exerted control on the Cenozoic global oceanic Sr isotope budget [Edmond, 1992]. Erosional inputs from similar metamorphic complexes may have been involved in previous major orogenies and related excursions of the Sr isotope curve by supplying Sr with extreme isotopic composition to the ocean.
 The evolution of the Os isotopes provides complementary insight into the processes that control the isotope budget of long residence time tracers in the ocean. In the case of Os, which has a much shorter oceanic residence time than Sr, there is some evidence for isotopic variations on short timescales (10,000 years) extracted from metalliferous sediments. These variations have been attributed to changes in weathering regime on glacial/interglacial timescales [Oxburgh, 1998] and during the Paleocene thermal maximum [Ravizza et al., 2001]. Long-term Os isotope records have been obtained for the past 70 Myr [Pegram et al., 1992; Ravizza, 1993; Peucker-Ehrenbrink et al., 1995; Pegram and Turekian, 1999] (see also the recent review by Peucker-Ehrenbrink and Ravizza ) (Figure 6c). The Os isotope record looks similar to the Sr isotope evolution in that it shows a general increase between about 65 Ma and present. In contrast to the Sr isotope record, however, the Os isotopes show high values prior to 65 Ma, and there have been several excursions toward lower values in the Cenozoic, the most pronounced of which occurred at the Eocene/Oligocene boundary at about 33 Ma (see Peucker-Ehrenbrink and Ravizza  for discussion of the timing). The major minimum at the Cretaceaous/Tertiary boundary has been interpreted as a consequence of the impact of a large extraterrestrial body which supplied significant amounts of Os with low 187Os/188Os. The slow increase in the Os isotope ratio after the impact may be explained by weathering of impact material on the continents [Peucker-Ehrenbrink et al., 1995]. The minimum at the Eocene/Oligocene boundary is less well constrained and occurred only in one core. An extraterrestrial origin cannot be excluded, but it is also possible that this feature is a leaching artifact of volcanic ash which requires verification by further records.
 Comparison of the general trend of the Os and Sr isotope records suggests that related processes have controlled the oceanic isotope budget for both elements. However, evidence has recently been presented that the changes in Os isotopes are unlikely to have originated from Himalayan riverine supply, because the dissolved Os concentrations are too low to have such a large impact on the Os isotope composition of the ocean [Sharma et al., 1999]. This suggests that Himalayan weathering and erosion has probably not been the only factor controlling the global oceanic budget of both isotope systems during the Tertiary and that more global variations in weathering regime and intensity have been involved.
5. ISOTOPIC EVOLUTION OF SHORT AND INTERMEDIATE RESIDENCE TIME TRACE METALS (Nd, Hf, Pb, Be): IMPLICATIONS FOR PALEOCEANOGRAPHY AND CONTINENTAL WEATHERING HISTORY
 The deep-water isotopic composition of short to intermediate residence time trace metals in the ocean (Nd, Hf, Pb, Be) is recorded by authigenic chemical precipitates in the ocean without significant fractionation processes. This is an important difference compared with classic nutrient-related isotopic water mass proxies, such as carbon isotopes. If a reliable age control for the authigenic deposits is available, the trace metal isotope composition of the ambient deep water in the past can be reconstructed.
 There have been attempts to extract the Nd isotope composition from the calcite of planktonic foraminifera as primary authigenic phase in the ocean, which contains only very low amounts of Nd (only few parts per million at maximum). There has been a debate as to the validity of these measurements due to potentially incomplete separation from postdepositional Mn-oxyhydroxide coatings, which contain much higher Nd concentrations than the foraminiferal calcite itself. With the advance of more reliable cleaning techniques, it has recently been suggested that the surface water Nd isotope composition can be extracted reliably from the calcite of planktonic foraminifera deposited in marine sediments. Such a record was obtained in the Labrador Sea for the past 2.5 Myr [Vance and Burton, 1999]. For the past 150 kyr, variations in a surface water Nd isotope record obtained from planktonic foraminfera from the Bay of Bengal were interpreted as changes in riverine inputs of Nd as a function of monsoonal circulation [Burton and Vance, 2000]. Although there are still uncertainties concerning the true Nd concentration of foraminiferal calcite prior to coating with the Mn-oxyhydroxides in the sediment [Pomies and Davies, 2000], which will require future studies, Nd isotopes in foraminferal calcite are a promising new tool for continuous reconstructions of surface ocean paleoceanography.
 For the present-day Atlantic Ocean [Palmer and Elderfield, 1985] (Figure 9a) and for the past 60 Myr at a location on the Rio Grande Rise in the South Atlantic [Palmer and Elderfield, 1986] (Figure 11a), the deep-water Nd isotope composition has been reconstructed from the early diagenetic ferromanganese coatings of foraminifera. Potential problems are caused by the diagenetic mobility of Mn and REE in the sediments under reducing conditions. The coatings may not have only derived their Nd from the bottom water but may also contain contributions from pore waters, which may have a different isotope composition and may bias a true deep-water isotope signal [Elderfield et al., 1981]. More recently, glacial-interglacial Nd isotope records obtained from the Mn coatings of South Atlantic sediment cores, which have obviously remained oxidized for the past up to 70 kyr, were used to reconstruct the NADW export into the Southern Ocean [Rutberg et al., 2000; Bayon et al., 2002] (section 5.3.2 and Figure 15).
 Reconstructions of the origin, strength, and flow paths of paleocurrents with radiogenic isotopes (Sr, Nd, Pb) have not only been derived from the dissolved water column signal but have also utilized the isotopic signatures of the detrital fraction carried by currents, which is ultimately deposited in marine sediments. It is obvious that the clay mineral fraction is most suitable for the reconstruction of certain water masses because of its slow sinking speed and thus long residence time within a water mass, whereas the larger grain size fractions have more likely recorded the detrital supplies from nearby outcropping terranes [Innocent et al., 2000]. Such studies have mainly focused on glacial-interglacial variations of bottom waters in the North Atlantic region where distinct isotopic differences of the source regions allow a detailed reconstruction of sources and mixing of different water masses [Grousset et al., 1988; Revel et al., 1996; Bout-Roumazeilles et al., 1998; Hemming et al., 1998; Fagel et al., 1999, Innocent et al., 2000], but there is also a study of the source provenances and circulation patterns in the Indian Ocean using Nd isotopes in the detrital fraction of marine sediment [Dia et al., 1992].
5.1. Ferromanganese Crusts: Genesis and Age Dating
 Recently, most attention has been devoted to extracting continuous radiogenic isotope time series from depth profiles of hydrogenous ferromanganese crusts which grow directly from the water column on outcrops of hard substrate such as hyaloclastite or basalt on the seafloor. The requirement for growth of crusts is the absence of particle sedimentation, which either occurs through bottom currents or steep slopes, for example, on seamounts (Figure 7). In addition, depth profiles of hydrogenous Mn nodules which grow on top of pelagic sediments in areas of very low sedimentation rates have been used. If these nodules get buried in oxygenated marine sediments, they preserve their radiogenic isotope composition and can be used to some extent for time series reconstructions on the megayear scale using the sediment stratigraphy [Winter et al., 1997; Ito et al., 1998].
 Trace metals are incorporated into crusts and nodules by a coprecipitation process. Although the exact mechanism of crust precipitation is unclear, it probably involves a two-stage process [Koschinsky and Halbach, 1995]. First, Mn and Fe from the water column form mixed colloids which scavenge trace metals by surface sorption. Second, these colloids precipitate together with the sorbed trace metals as amorphous oxide or oxihydroxide encrustations on any substrates on the seafloor. This process, in combination with the very slow growth rates, enriches the concentrations of trace metals in the crusts and nodules by up to 10 orders of magnitude compared with the seawater concentrations.
 It has recently been confirmed that isotope records of highly particle reactive elements (Th, Be, Pb, Nd) obtained from the highly porous ferromanganese crusts dating back millions of years are not significantly affected by diagenetic alterations or postdepositional exchange with present-day seawater. Henderson and Burton  calculated the rates of postdepositional exchange of trace metals in crusts from a comparison with the mobility of U. For Hf and Os their calculated exchange rates are too high owing to erroneous seawater concentrations. Using the correct seawater concentrations for Hf [McKelvey and Orians, 1998] and Os [Levasseur et al., 1998], the (much lower) exchange rates indicate that long-term records of Hf and Os isotopes in ferromanganese crusts can also be considered reliable. Less particle reactive elements (U, Sr, Li) with long oceanic residence times are not reliable because of exchange processes with seawater after deposition. This prevents a direct comparison of time series of Sr isotopes with those of other isotopes in crusts and makes the Sr isotope dating of crusts impossible. Further evidence for the reliability of radiogenic isotope records in crusts comes from the existence of coherent and sharp changes in the high-resolution Pb isotope composition of even 30- to 50-Myr-old sections of ferromanganese crusts D11-1 and CD29-2 in the Pacific Ocean, which are separated by more than 3000 km [Christensen et al., 1997]. These changes cannot be correlated to a major period of phosphatization, confirming the robustness of the crust records. An additional argument is that datings of the same crusts obtained from profiles of different radioactive trace metals such as 230Th or 10Be give consistent results (see Frank et al. [1999a] for further discussion).
5.2. Present-Day Deep-Water Isotope Composition From Ferromanganese Crust Surfaces
 Superimposed on the general provinciality are patterns ascribed to ocean circulation. In the case of Be the isotopic provinciality is almost exclusively a function of the nutrient-like behavior of 10Be which gets enriched along the flow of deep water from the Atlantic into the Pacific [von Blanckenburg et al., 1996a]. For Nd and Pb isotopes a tongue of low ϵNd (−9 to −11) and high 206Pb/204Pb (18.85 to 19.07) in the eastern sector of the Southern Atlantic has been attributed to the advection of NADW [Albarède and Goldstein, 1992; Abouchami and Goldstein, 1995; Albarède et al., 1997] (Figure 9). In the case of 206Pb/204Pb the NADW advection is apparently trackable into the SW Indian Ocean by high ratios (Figure 9b) [Vlastélic et al., 2001], whereas there is no indication for this in the Nd isotope data [Albarède et al., 1997]. Similarly, a tongue of low 206Pb/204Pb (18.7 to 18.8) and high ϵNd (−5 to −7) has been observed in the western sector of the Southern Atlantic and has been interpreted as inflowing Pacific waters. It was noted, however, that this is also the only major area where available deep-water ϵNd data do not agree with the nodule surface data, which was attributed to an integrated effect of decreased glacial NADW advection [Albarède et al., 1997] (see also section 5.3.2). In the Pacific Ocean a gradual northward increase in ϵNd [Albarède and Goldstein, 1992] and a clear relationship between water depth and Nd isotope composition at least for the central Pacific Ocean [Aplin et al., 1986/Aplin et al.,1987] are observed, which mirrors the northward advection of AABW with its lower ϵNd signature [Piepgras and Jacobsen, 1988; Shimizu et al., 1994]. Similarly, there is a reflection of the isotopic differences of the prevailing water masses and thus also water depth in the ϵNd signature of crusts from the Atlantic Ocean; for instance, crusts having grown from NADW have a lower ϵNd compared with those having grown from deeper waters (AABW) or shallower waters (AAIW) than the depth range of NADW (∼ 1500–3500 m). A similar relationship between water depth and Pb isotope distribution in crusts from the Indian Ocean [Vlastélic et al., 2001] and the Pacific [Abouchami and Galer, 1998] has also been suggested to be a consequence of isotopic differences between water masses.
 The provinciality of the isotopic signatures has been most stable for Nd isotopes (Figure 11a), which only show an overlap between the data from the Atlantic sector of the Southern Ocean and the two shallowest crusts from the northern Atlantic Ocean for the period between 5 and 8 Ma. In the case of crust 65GTV (1500 m water depth) this reflects the continuous influence of Mediterranean Outflow Water (MOW) with its typical ϵNd value around −9.4 [Spivack and Wasserburg, 1988; Abouchami et al., 1999]. The high Nd isotope values of crust BM1963.897 from the Blake Plateau (850 m water depth) for the past 5 Myr have been caused by admixture of water masses with relatively high ϵNd values from the Caribbean or from the dissolution of average continental dust, whereas the even higher values prior to 5 Ma probably reflect a contribution of surface and intermediate water masses from the Pacific Ocean prior to the closure of the Panama gateway (Figure 12)[Reynolds et al., 1999; Frank et al., 1999b] (see section 5.3.3). The records from the South Atlantic are in good agreement with a Nd isotope time series from the Rio Grande Rise obtained from ferromanganese oxide coatings [Palmer and Elderfield, 1986] (Figure 11a) and the central equatorial Pacific Ocean records agree well with Nd isotope records obtained from fish teeth [Martin and Haley, 2000]. The data reconstructed from the crusts are not, however, in good agreement with the reconstruction of the Nd isotope composition of Atlantic seawater as obtained from a combination of other discontinuous reservoirs such as phosphates, carbonates, or fish remains [Shaw and Wasserburg, 1985; Staudigel et al., 1985; Stille, 1992; Stille et al., 1996]. In view of the presently observed Nd isotope distribution in the Atlantic Ocean and its close relation to water depth and ocean water masses, this is, however, not surprising because, mostly, there is no information on water depth and thus potential water masses for these different locations [Stille, 1992; Stille et al., 1996].
Table 2. Locations of Crusts With Published Radiogenic Isotope Time Series
Average growth rates were derived from 10Be/9Be ratios.
In cases where the ages of the bases of the crusts exceed 7–8 Myr, the maximum ages were calculated by extrapolating the growth rates derived from 10Be concentrations in the upper sections of crust 121DK [Abouchami et al., 1999] or by using Co chronometry for the others [Frank et al., 1999a].
 The Hf and Be isotope records show considerable overlap between the Indian/Southern Ocean and Pacific Ocean records, whereas the data of the Atlantic Ocean and the Southern/Indian Ocean have stayed distinct from each other (Figures 11c and 11d). In the case of Be this is mainly explainable by the distinctly lower age of the Atlantic water masses, whereas for Hf the provenances of the source rocks combined with the difference in weathering regime in the North Atlantic area are mainly responsible for the persistent isotopic difference.
 All four isotope systems have in common, however, that there has been no overlap between the isotopic composition of the Atlantic and Pacific Oceans, although in the case of Pb isotopes, the values came quite close to each other for the periods between 50 and 60 Ma. This is probably the consequence of a more efficient mixing between the Atlantic and Pacific Oceans through the open seaway between North and South America at that time (see also section 5.3.5).
 Superimposed on the general provinciality, the isotope records of Nd, Hf, and Pb show pronounced trends over time, in particular over the past 15 Myr, some of which correlate basin-wide. The radiogenic isotope variations of the deep water of the past 60 Myr have thus predominantly been caused by either paleocirculation changes of the ocean, variations of inputs from the continents through weathering or provenance changes, or a combination of both and are discussed in more detail in the following sections (5.3.1–5.3.5). In general, the records show that the past 1 Myr has been the period of time with the most pronounced isotopic differences 2between the major ocean basins over the past 60 Myr. This is mainly a consequence of the distinct change in the isotopic composition of deep waters in the North Atlantic and will be discussed in more detail below (section 5.3.1).
 The influences of changes in ocean circulation and weathering processes have to be distinguished to facilitate the use of the radiogenic isotopes as paleocean tracers. Changes in weathering intensity and continental erosion have been the main factors controlling the long-term evolution of Sr and Os isotopes in the ocean, and it is thus obvious to compare these records to those of the potential paleocean tracers (Figures 6 and 11). It is immediately obvious that there are no clear relationships between them, which may have pointed to a common origin of the variations. For instance, in none of the Nd, Hf, or Pb isotope records can a long-term trend starting about 30 Ma be observed. The only records that resemble the Sr and Os records are the Pacific Nd isotope time series, which, however, have not continued to increase until the present day but started to decrease again for the past 3–5 Myr. If Himalayan erosion had played a major role for the Nd isotope evolution of the Pacific as has been suggested for Sr isotopes, the trend should point in the opposite direction because the Nd isotope signal of the eroded Himalayan material is very negative (ϵNd ∼ −16; see sections 5.3.4 and 5.3.5). The only clear indication for an influence of Himalayan erosion on deep-water radiogenic isotope composition is found in the local 208Pb/206Pb evolution of the deep Central Indian Ocean (crust SS663, see section 5.3.4). It thus seems that the global trends in erosional input and weathering regime are not reflected by the Nd, Hf, and Pb isotope composition of the deep ocean. Clearly, more regional changes in either weathering and erosion processes or water mass circulation have been dominant, which is a consequence of the shorter oceanic residence time of Nd, Hf, and Pb compared with that of Os and Sr. In sections 5.3.1–5.3.5, results obtained from ferromanganese crusts for which a distinction between erosional/weathering processes and variations in ocean circulation has been possible are discussed.
5.3.1. Radiogenic isotope evolution of NADW in the North Atlantic Ocean
 NADW is produced in the North Atlantic Ocean, which makes this area one of the key locations controlling the global thermohaline circulation system and global climate. In the western North Atlantic, three crusts located within the flow path of present-day NADW (BM1969.05, ALV539, TR079) have recorded its radiogenic isotope evolution over the past up to 60 Myr (Figures 11 and 13). The most significant feature of these records is a relatively constant Pb and Nd isotope composition prior to about 3 Ma and a strong trend toward lower Nd isotope, 207Pb/206Pb, and 208Pb/206Pb ratios and higher 206Pb/204Pb ratios afterward [Burton et al., 1997, 1999a; O'Nions et al., 1998; Reynolds et al., 1999]. In the eastern Atlantic basin a similar pattern with a smaller amplitude (crust 121DK) has been ascribed to advection of NADW into the eastern Atlantic basin [Abouchami et al., 1999]. A comparable decrease in the Hf isotope ratios in western North Atlantic crusts (BM1969.05 and ALV539) has been found over the same period [Piotrowski et al., 2000; van de Flierdt et al., 2002]. Prior to this the Hf isotopes in the one long record available for the western North Atlantic (BM1969.05) show a trough-like undulation starting at about 30 Ma with a broad minimum between ∼12 and 23 Ma, which is not paralleled by the Pb and Nd isotope records and is not yet explained. From the patterns of these deep-water Nd, Hf, and Pb isotope records alone, it is not possible to decide whether changes in ocean circulation or changes in weathering regime on the continents have been the main cause for the observed variations over the past 3 Myr.
 Initially, the Nd isotope variations of NADW over the past 3 Myr (Figures 11, 12, and 13) were explained by an increasing contribution of deep water formed in the Labrador Sea [Burton et al., 1997], which at present has a high Nd concentration and a very low ϵNd of –22, caused by weathering of Archean continental crust with values as low as −40 [Stordal and Wasserburg, 1986; Hemming et al., 1998]. Consequently, even a small increase in the contribution of Labrador Seawater would lead to a decreased Nd isotope signal in NADW. In addition, it was initially argued that the final closure of the Panama gateway at about 3.5 Ma stopped the admixture of Pacific water masses with very radiogenic Nd (high ϵNd) and at the same time strengthened NADW production, thus also suppressing the supply of water masses with a relatively high ϵNd from the southern Atlantic into the North Atlantic [Burton et al., 1997]. These issues will be discussed in detail in sections 5.3.2 and 5.3.3, where it will be shown that it is unlikely that NADW production increased at the onset of Northern Hemisphere glaciation and that the closure of the Panama seaway for significant water mass exchange had already occurred at about 4.6–5 Ma.
 Recently, it has been shown that a change in weathering regime in the high northern latitudes was mainly responsible for the observed trace metal isotope changes. Nd isotope ratios have probably decreased owing to an increased amount of detrital input since the onset of Northern Hemisphere glaciation at about 2.7 Ma. Strongest evidence for a change in weathering regime comes from the Pb and Hf isotope records. Von Blanckenburg and Nägler  demonstrated that the Pb isotope composition of dissolved Pb in the North Atlantic and Arctic Ocean has been dominated by the preferential release of radiogenic Pb (206Pb, 207Pb, 208Pb) during weathering of rocks of the old continental shields of Canada and Greenland. Intense physical erosion such as during Northern Hemisphere glaciation favors the release of radiogenic Pb because of the continuous exposure of new rock surfaces and thus supposedly caused the observed variations in deep-water Pb isotope composition over the past 3 Myr (Figures 11b and 13). These processes are thus also responsible for the generally strong negative correlation between the ϵNd and 206Pb/204Pb time series and positive correlation between ϵNd and 207Pb/206Pb and 208Pb/206Pb in the Atlantic Ocean [Frank and O'Nions, 1998; Frank et al., 1999a]. In contrast, the sign of correlation between the Nd and Pb isotopes is opposite in the equatorial Pacific Ocean, probably because there are no pronounced incongruent weathering effects on Pb isotopes in this regime of predominant chemical weathering. The dissolved Nd and Pb isotope composition in the Pacific Ocean more closely reflects the input sources such as the island arcs (section 5.3.5).
 Independent evidence corroborating the importance of a change in weathering regime in the North Atlantic after 3 Ma comes from the almost concomitant decrease in Hf isotope composition of NADW (Figure 11c). This decrease can only be explained by an enhanced contribution of unradiogenic (low ϵHf) Hf hosted by zircons, which release some of their Hf during mechanical breakdown caused by enhanced physical erosion [van de Flierdt et al., 2002]. The products of the enhanced mechanical weathering (with low ϵHf and ϵNd as well as radiogenic Pb isotope ratios) were supplied to NADW via the Baffin Bay by Labrador Sea water [Vance and Burton, 1999]. Additional support for a weathering-based explanation comes from Mn micronodules incorporated into sediment cores from the Arctic Ocean which have recorded variations in Nd and Pb isotopes of Arctic Ocean deep water over the past 5 Myr [Winter et al., 1997]. These variations very closely resemble those of NADW but cannot have been caused by circulation changes in the Atlantic Ocean.
 In view of the pronounced change in the Nd, Hf, and Pb isotope composition over the past 2–3 Myr, one may expect a similar change in the 10Be/9Be ratios. The reason for this would be the increased input of 9Be into the ocean caused by enhanced weathering of the continental crust, whereas 10Be, which is supplied from the atmosphere, would at the same time remain unaffected. Such a change in the decay-corrected initial 10Be/9Be profiles is not observed (Figure 11d), but the ratios have remained constant within a small range [von Blanckenburg and O'Nions, 1999]. This appears to be an argument against the control of the observed patterns in the Nd isotopes by weathering [Burton et al., 1999a]. Further results indicate, however, that the surface water Nd isotope composition of the Labrador Sea experienced a similar general trend over the past 2.5 Myr as the deep NW Atlantic crusts [Vance and Burton, 1999]. It is thus more likely that the changes in detrital 9Be supply were too small to cause a resolvable change in the deep-water 10Be/9Be ratio of the North Atlantic [von Blanckenburg and O'Nions, 1999; von Blanckenburg and Nägler, 2001]. Thus the constant initial 10Be/9Be ratios argue against significant overall circulation changes in the North Atlantic over the past 3 Myr but do not exclude changes on glacial-interglacial timescales, which cannot be resolved by the crust records.
 In summary, these lines of evidence strongly suggest that an increased input of detrital material released by enhanced physical erosion of the old continental landmasses of the Canadian Shield and Greenland associated with the onset of Northern Hemisphere glaciation was the most important factor controlling the changes in isotope composition of deep waters in both the North Atlantic and Arctic Ocean over the past 3 Myr.
5.3.2. History of NADW Export Into the Southern Ocean
 From the reconstruction of the isotopic evolution of NADW in the North Atlantic alone, one cannot assess the strength of NADW production and the vigor of the global thermohaline circulation in the past. Analogous to approaches using δ13C of benthic foraminifera (see section 1.2), an estimate of NADW export can be achieved from a comparison of the Nd isotope evolution of NADW and the Southern Ocean. The strength of the admixture of low ϵNd NADW to the Southern Ocean over time should be reflected by changes of the Southern Ocean deep-water Nd isotope composition because presently about 50% of the ACC water masses are derived from NADW.
 Nd isotope time series of five crusts in the South Atlantic and the Southern Ocean, however, display no resolvable variability over the past up to 14 Myr. They do not reflect the pronounced changes in Nd isotope composition over the past 3 Myr observed in the western North Atlantic [Frank et al., 2002] (Figure 14). A Nd isotope record obtained from ferromanganese coatings of foraminifera from a water depth of 2050 m on the Rio Grande Rise confirms this observation by showing the same range of values as the Southern Atlantic crust records over the past 14 Myr [Palmer and Elderfield, 1985] (Figure 11a). In addition, two glacial/interglacial records of Nd isotope variations obtained the same way at two locations in the deep Cape Basin [Rutberg et al., 2000] average exactly at the same values as the crust data from corresponding locations. This lends further support to the reliability of reconstructions of the deep-water radiogenic isotope composition from the authigenic fraction of marine sediments despite potential problems of Nd contributions from pore waters and the diagenetic mobility of Mn under reducing conditions.
 In view of the well-constrained export of low ϵNd NADW in the present-day Atlantic Ocean (Figure 5), the most likely explanation for these observations is a progressive decrease in overall NADW export into the Southern Ocean since the onset of Northern Hemisphere glaciation at about 3 Ma. This obviously coincided with the decrease in the ϵNd signature of NADW, which was caused by increased weathering intensity and detrital input into the North Atlantic in the course of Northern Hemisphere glaciation (section 5.3.1). Pb isotope data obtained from the same crusts similarly indicate a decrease in the export of typical NADW Pb isotope signature to the Southern Ocean over the past 3 Myr. This is in agreement with previous interpretations of the distribution of Pb isotopes in Mn-nodule surfaces [Abouchami and Goldstein, 1995; Vlastélic et al., 2001] and with modeling results [Henderson and Maier-Reimer, 2002] suggesting that, indeed, a significant NADW Pb isotope signature can be advected from the North Atlantic into the Southern Ocean. The short oceanic residence time of Pb and the strong influence of local contributions from continental weathering, however, prevent any quantitative estimates of the decrease in NADW export (Figure 14).
 A decrease in NADW export during the glacial periods of the late Quaternary has been identified from stable carbon isotope comparisons of North Atlantic and Pacific records [Raymo et al., 1990,Raymo et al., 1992]. Nd isotope data recovered from the ferromanganese coating of foraminifera of glacial/interglacial Cape Basin sediments have recently been shown to reflect a greatly diminished NADW input into the Southern Ocean during glacial times [Rutberg et al., 2000] (Figure 15). The results of this Nd isotope record are entirely consistent with the carbon isotope record of the same sediment core, which also indicates a decreased glacial export of NADW [Charles and Fairbanks, 1992; Charles et al., 1996]. The Nd isotope record of this core is probably one of the most striking examples so far for the potential offered by radiogenic trace metal isotopes to reconstruct paleocirculation patterns.
 The new Nd isotope data from the Southern Ocean crusts now allow a semiquantitative estimate of the overall exchange between NADW and the main Southern Ocean deep-water mass CDW over the past 14 Myr. Such semiquantitative estimates are a valuable complementary addition to the carbon isotope approaches. Basic assumptions for this estimate are a NADW-free Pacific-dominated end-member ϵNd of Southern Ocean deep water in the range of −4 to −6 and no systematic differences in Nd concentration between NADW and CDW. Simple mass balance of the Atlantic and Southern Ocean Nd isotope data then suggests that at the Walvis Ridge (location of crust Atlantis II), 45–55% of the water presently consists of NADW, which is in good agreement with estimates based on oceanographic [Döös, 1995] and water column Nd isotope data from the Southern Ocean [Piepgras and Wasserburg, 1982]. The time series data (Figures 11, 14) indicate that the overall contribution of NADW to Southern Ocean water masses was between 15 and 35% higher at 3.5 Ma compared with the present [Frank et al., 2002]. It is noted that this estimate of decreasing overall NADW export is also consistent with decreased NADW export during the glacial periods of the past 3 Myr because the ferromanganese crust samples always integrate over several glacial-interglacial cycles [Albarède et al., 1997]. For the period between 14 Ma and 3.5 Ma the data indicate a continuous and strong export of a NADW-type water mass from the North Atlantic.
5.3.3. Influence of the Panama Gateway Closure on Ocean Circulation
 Paleogeographic reorganizations either initiate or prevent water mass exchange between ocean basins with different trace metal isotope characteristics and should thus be mirrored by changes in deep-water isotope composition of the corresponding ocean basins. The most recent important paleogeographic event was the closure of the Panama gateway by the emergence of a land bridge between North and South America at ∼3.5 Ma [Keigwin, 1982]. There is considerable debate as to when direct deep-water exchange between the Atlantic and Pacific Oceans was stopped because the paleogeography of the Carribbean region is complicated by the presence of several tectonic microplates. There is evidence for a shallow sill (1000 m) at 12 Ma [Duque-Caro, 1990] and a land bridge as early as 8 Ma [Marshall, 1985]. Faunal comparisons between Pacific and Caribbean deep-dwelling microorganisms suggest that the evolution of species in both basins did not occur in parallel after about 4.6 Ma. This suggests a cessation of deep-water exchange between the two oceans and is in agreement with sedimentological results obtained from Caribbean and Pacific carbonate sediments [Haug and Tiedemann, 1998]. The consequences of the closure of the Panama gateway for the global thermohaline circulation and global climate have also been subject to divergent opinions. A causal relationship between closure of the Panama gateway and the approximately coinciding onset of Northern Hemisphere glaciation was inferred [Stanley, 1995], whereas others suggested that the Panama closure for deep-water exchange as early as 6 Ma inhibited the onset of Northern Hemisphere glaciation for several million years [Berger and Wefer, 1996]. Sedimentological results support an increase in NADW production rate coinciding with the closure for deep-water exchange at about 4.6 Ma [Haug and Tiedemann, 1998] and are in agreement with results obtained from carbon isotope comparisons of the deep Pacific and Atlantic basins [Raymo et al., 1992; Ravelo and Andreasen, 2000].
 Reconstructions based on trace metal isotope records in ferromanganese crusts have recently shed new light on the consequences of Panama gateway closure for ocean circulation. Pacific deep-water crust (GMAT 14D), which recorded the Nd and Pb isotope composition close to Panama over the past 7 Myr, shows only a very small shift toward higher ϵNd values at 4.5 Ma and no change in Pb isotope composition [Frank et al., 1999b] (Figure 13). Assuming this shift in Nd isotope values was caused by a stop of the admixture of North Atlantic Water masses across Panama, mass balance suggests that less than 5% Atlantic water contributed to Pacific deep water during the 3–4 Myr prior to complete closure of the Panama gateway.
 On the Atlantic side of Panama, a crust from a water depth of 850 m on Blake Plateau (BM1963.897) is now bathed in Gulf Stream water. This location was particularly sensitive to circulation changes induced by the closure of the Panama gateway because of the stopped inflow of shallower subsurface water masses with Pacific Nd (ϵNd −2 to −4) and Pb (206Pb/204Pb ∼ 18.7) isotopic composition into the Atlantic. In addition, northward deviation of surface water masses originating from the South Atlantic which would have flowed into the Pacific prior to closure would result in similar changes for the Nd and Pb isotope composition of water masses on Blake Plateau. The isotope time series of the Blake Plateau crust shows two significant patterns for both Pb and Nd [Reynolds et al., 1999]. At about 2–3 Ma the Nd and Pb isotopes show a shift comparable to the deeper western North Atlantic crusts [Burton et al., 1997, 1999; O'Nions et al., 1998; Reynolds et al., 1999] (Figure 13). Following the interpretation given in section 5.3.1, this is probably a consequence of within-basin mixing of NADW, which is, for example, also evident in similar isotope patterns in two crusts from the eastern Atlantic Basin [Abouchami et al., 1999].
 Prior to 3 Ma the water mass on the Blake Plateau shows an isotopic evolution that is not observed in other crusts located farther away from the Panama gateway in the North Atlantic. Over the period between ∼8 and 5 Ma the Nd isotope composition decreased by 2.5 ϵNd units and 206Pb/204Pb increased from 18.83 to 18.97, with the change for the Pb isotopes having already been completed by 6.5 Ma. These patterns are interpreted as a progressive decrease in the advection of Pacific water masses into the Atlantic Ocean. The data suggest that the Panama gateway was already closed to deep and intermediate water mass exchange by ∼5 Ma, in agreement with sedimentological evidence [Haug and Tiedemann, 1998]. This implies that the closure of the Panama gateway cannot have been the direct cause but rather only one of the prerequisites for the onset of the Northern Hemisphere glaciation [Reynolds et al., 1999]. The immediate consequence of the Panama gateway closure for deep-water exchange at 4.6 Ma was thus early Pliocene climate warming rather than a major cooling.
 In summary, the radiogenic isotope record from the Blake Plateau is a good example for the potential of the radiogenic isotope records in paleoceanographic research in that it first documented changes in ocean circulation for the period between 8 and 5 Ma, and then from 3 Ma to present it has been influenced by changes of the weathering regime, similar to other crusts from different water depths in the North Atlantic basin.
5.3.4. Radiogenic Isotope Evolution of Indian Ocean Deep Water and the Influence of Himalayan Uplift
 The Indian Ocean has been the sink for erosional products of Himalayan uplift over the past 30 Myr, as is evidenced by the sedimentary deposits of the Bengal and Indus fans. It may thus be expected that the dissolved trace metal isotopic composition of the Indian Ocean has also been influenced by Himalayan erosion products over this period of time as is suggested by Sr isotopes. An erosional signal from the Himalayas should be particularly pronounced during times of strongest uplift and thus maximum erosion, which probably started around 24 Ma [Searle, 1996].
 Nd isotope and 206Pb/204Pb records of crust SS663 from the deep Central Indian Ocean are compared with other crust records in Figure 11 [O'Nions et al., 1998; Frank and O'Nions, 1998]. Over such long periods of time, factors such as subsidence and plate tectonic drift of the locations of the crusts have to be considered as potential causes for the observed isotopic trends of the ambient deep water. The location of crust SS663 from the central Indian Ocean has subsided by 2 km water depth from a location above the paleo-CCD in the Eocene to a water depth below the CCD at present. Growth of a ferromanganese crust at this location only started about 26 Ma, probably as a consequence of a higher deep-water oxygenation and decreased detrital sedimentation [Banakar and Hein, 2000]. In addition, the location of the crust moved about 2200 km northeast from its initial location during the past 53 Myr. These horizontal and vertical movements are obviously reflected by changes in the supply of detrital particles, growth structures, and major element geochemical composition of crust SS663 but have apparently not left any clear signal in its radiogenic isotope record.
 One of the most striking observations of the time series of this crust is the constancy of the Nd isotope composition in contrast to the marked change in Pb isotopes (Figures 11a and 11b). The Nd isotope data throughout this crust have been within the narrow range of present-day Indian Ocean deep water (ϵNd = −7.4 to −8) [Bertram and Elderfield, 1993; Albarède et al., 1997] for the past 26 Myr and suggest that the huge input of Himalayan weathering material with an ϵNd value of –16±2 [Derry and France-Lanord, 1996] did not significantly change the Nd isotope budget of the Indian Ocean deep water. In contrast, there was an increase of the 206Pb/204Pb from a low of 18.6 (similar to the deep Pacific) to a high around 18.9 at present. While this range is still intermediate between the Atlantic and Pacific Ocean data, an even more pronounced signal is observed for the time series of the 208Pb/206Pb ratio, which peaked between 20 and 8 Ma (Figure 16) at values even higher than the present Pacific. This signature has been attributed to Himalayan erosional input [Frank and O'Nions, 1998] mainly derived from the High Himalayan gneisses and leucogranites [France-Lanord et al., 1993]. This is supported by an offset of the Pb isotope data of the Central Indian Ocean toward the Pb isotope composition of these rocks in Pb-Pb isotope space (Figure 17). This observation demonstrates the potential of the three radiogenic Pb isotopes that are derived from different sources and thus provide independent information on input processes and sources.
 The reason for the apparent decoupling of the seawater Pb and Nd isotope responses to Himalayan erosion remains somewhat speculative. One possible reason is that the Nd is retained more efficiently in the estuary of the Ganges and the sediments of the Bengal Fan than is Pb. An alternative explanation is that most of the Pb introduced by rivers into the Indian Ocean through Himalayan erosion is scavenged and deposited more efficiently than Nd proximal to the riverine sources, i.e., at the location of SS663. Nd is much less particle reactive, and the riverine Himalayan Nd may therefore be efficiently mixed with other water masses of different Nd isotope composition and advected into other ocean basins so that no particular signal of Himalayan erosion has been left in the deep northern Indian Ocean.
5.3.5. Radiogenic isotope evolution of the equatorial pacific deep water
 The equatorial Pacific Ocean is another example of the competition between ocean circulation changes and weathering inputs in controlling deep-water radiogenic isotope composition. Nd and Pb isotope records (and, to a lesser extent, Hf isotopes) derived from equatorial Pacific ferromanganese crusts show well-resolved and comparable patterns over the past 50–60 Myr. For Nd, there is a more or less steady increase from ϵNd values between –5 and –4.5 to maximum values around –2.8 at about 3–5 Ma in the two seamount crusts CD29-2 and D11-1 from water depths around 2000 m (Figure 11a). The rate of change in Nd isotopes was strongest between about 15–12 Ma and 5 Ma. After 3–5 Ma the trend reversed toward slightly lower ϵNd values [Ling et al., 1997]. Applying the revised timescale given by Frank et al. [1999a], a third deep-water crust VA13/2 from the equatorial Pacific shows a very similar pattern over the past 26 Myr, but with a constant offset of about 1 ϵNd unit toward lower values, which is probably a reflection of a long-existing advection of AABW with its lower Nd isotope ratio into the deep central Pacific Ocean [Ling et al., 1997]. A similar difference in ϵHfbetween deep crust VA13/2 and the other equatorial Pacific crusts is probably also related to the influence of AABW [David et al., 2001]. A fourth crust, GMAT 14D, from deep waters close to Panama also shows a similar pattern as the others over the past 7 Myr but with a 1 ϵNd unit offset toward higher values (Figure 11a), providing evidence for some imprint of weathering products from the young nearby island arc rocks of Central America [Frank et al., 1999b].
 The patterns of the Pb isotope time series obtained for these crusts by different analytical techniques also show a generally close agreement [Ling et al., 1997; Christensen et al., 1997; Abouchami et al., 1997; Frank et al., 1999b; David et al., 2001] (Figure 11b). In particular, the patterns of the two high-resolution laser ablation MC-ICPMS Pb isotope profiles obtained on the two seamount crusts CD29-2 and D11-1 reveal close correspondence, despite the fact that they are separated by some 3000 km [Christensen et al., 1997] (Figure 16). The high-resolution variability of the Pb isotope ratios seems to correlate with the deep-water oxygen isotope record and thus global climate and ice volume. Christensen et al.  have also suggested that larger differences between the absolute values of the two Pb isotope records prior to about 30 Ma may indicate a weak circulation and water mass mixing in the Pacific basin for that period of time. After 30 Ma, Pb isotope records became very similar, indicating efficient mixing of Pb derived from various input sources in the Pacific Ocean despite its relatively short oceanic residence time (Figure 16). A similar interpretation was derived for Hf isotopes, which have apparently been very well mixed in the equatorial Pacific at water depths around 2000 m for the past 20 Myr [Lee et al., 1999] (Figure 11c).
 The similar patterns of all four crusts suggests some common control for the changes in the Nd and Pb isotope composition of the entire equatorial Pacific Ocean. The continuous increase in Nd isotope composition over the past 15 Myr, particularly between 10 and 5 Ma, approximately coincides with the onset of deep-water production (AABW) around the Antarctic. One might expect that this circulation change also influenced the trace metal isotope composition of deep-water masses in the Pacific. Whereas the relatively small change in the 206Pb/204Pb time series over this period of time (Figure 11b) would be consistent with an increased contribution of AABW, the increase in Nd isotopes clearly goes the opposite way, which excludes this possibility. A more likely source for the observed long-term trend in the dissolved Pacific Nd isotope composition appears to be the evolution of the Pacific island arcs such as Indonesia and the subsequent weathering of these rocks with their very high ϵNd signature. This is in agreement with the inferred evolution of the Pacific island arc production rate in the past, which was relatively low between 15 and 8 Ma and has increased strongly afterward [Kaiho and Saito, 1994]. Earlier periods of high arc production rates between 35 and 15 Ma are not reflected by Nd isotope records, possibly as a function of lower total amounts of weathering products supplied to the ocean over this period. It has been shown that today REEs from tropical island arcs are very efficiently introduced into Pacific surface waters by riverine input and remobilization on the shelves [Sholkovitz et al., 1999; Lacan and Jeandel, 2001]. For the dissolved Pb isotope composition of the Pacific, it has also been suggested that weathering of the Pacific island arcs may play an important role [von Blanckenburg et al., 1996b].
 The coinciding small decrease in ϵNd and changes in Pb isotope composition of Pacific deep waters at 3–5 Ma (Figures 11a, 11b, and 16) had (in the case of Nd) been attributed to advection of the pronounced isotopic changes within NADW (section 5.3.1) into the deep Pacific Ocean [Ling et al., 1997; Martin and Haley, 2000]. In view of the previously published records (nearly invariant Nd isotope record from the Indian sector of the Southern Ocean [O'Nions et al., 1998] and, in particular, the new data from the Southern Ocean discussed in section 5.3.2 [Frank et al., 2002]), this is not a viable explanation because the NADW signal should have left its imprint on the isotope composition of the deep Southern Ocean on its way to the deep Pacific. This is not the case, probably as a consequence of the inferred decrease in NADW advection to the Southern Ocean. For Pb the oceanic residence time is too short to transfer a signal from the North Atlantic into the Pacific [Henderson and Maier-Reimer, 2002]. Alternatively, the small change in Nd and Pb isotope composition might be explained by an increased inflow of Southern Ocean water masses [Abouchami et al., 1999].
 Given that there was no decrease in island arc production over the past 3–5 Myr [Kaiho and Saito, 1994], the most likely explanation for the changes in Nd and Pb isotopes seems to be a change in eolian inputs, which have been suggested to be important factors in controlling the dissolved radiogenic isotope signal in the ocean [Abouchami et al., 1997; Tachikawa et al., 1999]. This is a reasonable interpretation in view of the fact that the eolian input into the Pacific derived from Asia increased by about an order of magnitude around 3 Ma [Rea, 1994], coinciding with the changes in dissolved Nd and Pb isotope composition of the deep Pacific Ocean. Changes in eolian dust supply and deposition prior to 5 Ma have probably been too small to influence the dissolved radiogenic isotope budget of the Pacific Ocean.
 Over the up to 60 Myr covered by the equatorial Pacific crusts the plate tectonic movement of their locations relative to the Pacific climate and wind belts and current systems may have had an influence on their radiogenic isotope evolution. An equatorial Pacific crust which has moved from a position at the equator to 25°N over the past ∼80 Myr shows, for example, element compositional changes that have been related to the movement of the crust location relative to the Pacific trade wind zones [McMurtry et al., 1994]. The location of equatorial Pacific sediment core GPC3 has a similar paleopath and recorded clear Nd and Pb isotopic shifts of the deposited eolian dust caused by a change in dust source provenance from Northern America to Asia [Pettke et al., 2002]. Equatorial Pacific crusts with similar paleopaths, however, do not show clearly related shifts in their radiogenic isotope records [Ling et al., 1997; Christensen et al., 1997; Lee et al., 1999]. An increasing eolian influence from Asia should, for example, be reflected by a decrease in ϵNd over the past 20–30 Myr, whereas the opposite trend is observed.
 In summary, the available records from the equatorial Pacific Ocean suggest that changes in weathering sources and dust input rather than changes in ocean circulation have dominated the deep-water trace metal isotope composition over the past 30 Myr. Further records from the south and north Pacific are needed to verify this.
6. SUMMARY AND OUTLOOK
 Radiogenic isotope time series of the deep ocean provide important information for the reconstruction of paleocirculation patterns as a consequence of paleoclimatic or paleogeographic changes, changing inputs due to changes in weathering regimes, weathering intensity, or mountain uplift. These different factors have to be distinguished in order to extract reliable information on distinct processes. For the reconstruction of changes in regime and intensity of weathering, this has been successful in three example areas. In the deep North Atlantic, radiogenic isotope records have mainly varied as a function of weathering changes associated with the onset of Northern Hemisphere glaciation. In the Indian Ocean a Pb isotope time series has obviously recorded the intensity of Himalayan erosion. In the equatorial Pacific Ocean, changes in the erosional input associated with the evolution of the island arcs have mainly controlled the radiogenic isotope variability. In contrast, a shallow North Atlantic record from Blake Plateau documented changes in ocean circulation related to the closure of the Panama gateway, and a comparison of the radiogenic isotope evolution of the Southern Ocean with that of NADW documented a decrease in the intensity of NADW export to the Southern Ocean over the past 3 Myr. These and other studies have resulted in a greatly improved understanding of the main input sources and the processes controlling the oceanic distribution of the dissolved signatures originating from these sources.
 To reconstruct past circulation patterns of the oceans on timescales of millions of years, there are a number of interesting key areas from which ferromanganese crust–based studies could provide additional valuable information. Provided that suitable crust material can be found, this includes the timing and the oceanographic consequences of the opening of the Drake Passage or the closure of the Indonesian gateway, or the evolution of mass exchange through oceanic fracture zones of mid-ocean ridges, which should be highly sensitive to circulation changes. In addition, there is a lack of long-term Pacific time series which are not from equatorial regimes.
 With the advance of methods to extract the trace metal isotope information from the authigenic fraction of pelagic marine sediments, the future of radiogenic isotopes as paleoceanographic proxies lies mainly in applications to changes on much shorter timescales such as glacial/interglacial or even decadal or centennial instabilities of the ocean and their causes and responses to abrupt climate change. Pelagic sediments have the advantage of a much better time control and availability of study material from almost everywhere in the ocean, which should provide more quantitative estimates on past water mass exchange. In addition, information on changes in weathering regimes and erosional inputs can be extracted from the detrital fraction of the same material.
 Future work should also include a more detailed study of the input mechanisms in key areas (such as the Ganges and Brahmaputra estuaries or high-latitude rivers and estuaries), which are close to those of the most pronounced isotopic changes of the deep ocean. In addition, a more detailed study of the eolian input and the processes that make this material available for the dissolved radiogenic isotope budget of the ocean is needed in order to better quantify the importance of this source and its local distribution.
 Such parameterization will enable more detailed and better constrained modeling of the present and past oceanic behavior of radiogenic isotopes, which will improve the understanding of the relevant processes in the ocean. Together, such studies will allow better constrained applications of radiogenic isotopes as tracers in oceanography and paleoceanography.
 I would like to thank F. von Blanckenburg for numerous discussions about the use of trace metal isotopes in the ocean and for critical comments on the manuscript. R. K. O'Nions, A. N. Halliday, B. Reynolds, N. Whiteley, N. Belshaw, K. David, T. van de Flierdt, and J. R. Hein are thanked for many discussions. The paper benefited greatly from reviews of H. Elderfield and L. Derry and constructive comments of the editor T. Torgersen. I would like to thank the Swiss Science Foundation for support. Much of the work presented was carried out during support by a grant within the Training and Mobility through Research network program “The Marine Record of Continental Tectonics and Erosion” of the European Union.
 Tom Torgersen was the Editor responsible for this paper. He thanks one anonymous cross-disciplinary reviewer and one anonymous technical reviewer.