3.1. Controls on the Level of Dissolved Pt
 Concentrations of dissolved Pt ranged from below detection limit (0.1 pM) to 5.8 pM (Table S1). Ninety-four percent of our samples (n = 186) were distinctly low (<0.1–1.4 pM), and the outliers (1.5–5.8 pM) represent approximately 6% (n = 12) of the samples (Figure 2). Nine outliers (RD125, RD124, CJ239, CJ238, RD132, RD130, RD213, RD120, and CJ245) were extracted from the eastern Tibetan Plateau rivers (n = 138) and three (AU108, UL713, and IG106) from the rivers of the Russian Far East (n = 60). The concentrations of Pt in the six samples attributed to the springs of Mt. Baekdu (CB101–106) and in a spring in the Lena basin (UL406) ranged between below detection limit and 5.2 pM and are treated separately from the riverine data. Our measurements are on the lower side of the few reported measurements for rivers (see Introduction).
Figure 2. Box plot of dissolved riverine Pt concentrations in (a) China and Vietnam – the Salween, Mekong, Chang Jiang (Yangtze), Hong (Red), Huang He (Yellow), and Duman and hot springs of Mt. Baekdu and (b) the Russian Far East (the Amur, Lena, Yana, Indigirka, and Kolyma). n: number of samples analyzed. Stars indicate outliers (n = 12, 6% of the total number of the samples) and are labeled with the sample ID. UL406 is a spring sample.
Download figure to PowerPoint
 We carried out PCA on dissolved Pt and other major elements as a first attempt at drawing out the source characteristics of Pt in river waters. The source lithologies indicated by PCA are not expected to correlate with surface exposure areas within the sampled sub-basins due to incongruent weathering. For example, even a small exposure of carbonate can disproportionately affect the river dissolved load composition. Also, different elements in various minerals display diverse dissolution and adsorption/desorption kinetics. The results of PCA would represent the amalgamated effect of mineral type, dissolution kinetics, and exposure. PCA was performed separately for (i) the main group (n = 182, <0.1–1.4 pM Pt) and (ii) the outliers (n = 12, 1.5–5.8 pM Pt). For each calculation, two factors were extracted that explained 67% and 88%, respectively, of the variance in the two data sets (Table 2).
Table 2. Rotated Component Matrix for River Dissolved Pt and Major Elementsa
| ||Main (<1.4 pM Pt)||Outlier (1.5–5.8 pM Pt)|
|Factor 1||Factor 2||Factor 1||Factor 2|
|Cl−/Na|| ||−.599||.807|| |
|SO42−/Na||.674|| ||.940|| |
|Si/Na|| ||.950|| ||.918|
|Sr/Na||.696|| ||.913|| |
|% of Variance||38||29||46||42|
 For the main group samples, the Pt/Na was best grouped with Mg/Na, HNO3−/Na, Ca/Na, Sr/Na, SO42−/Na, and moderately with 87Sr/86Sr in the first factor, which can be interpreted as carbonate weathering (Table 2). The second factor, not associated with Pt, had high positive loadings for Si/Na and K/Na, a moderate negative loading for Cl/Na, and moderate positive loadings for 87Sr/86Sr, Ca/Na, and HNO3−/Na. The second factor may be due to biological activities such as phytolith growth and minor weathering of silicates. Negative loading for Cl/Na suggests dilution by halite dissolution. The variability in Pt concentrations for the main group samples is best associated with carbonate weathering. We do not put significant emphasis on this association, as it represents the average composition of the rivers dominated by carbonates and the low Pt concentrations due to low levels in the UCC and to low solubility in the absence of strong organic ligands.
 For the outlier samples the first extracted factor had high positive loadings for SO42−/Na, Sr/Na, Mg/Na, and Cl/Na and moderate loadings for HCO3−/Na, Ca/Na, Pt/Na, and K/Na. This factor can be related to evaporite and carbonate dissolution (Table 2). The second factor had high positive loadings for Si/Na, K/Na, 87Sr/86Sr, Pt/Na, and Ca/Na and moderate loadings for HCO3−/Na and Mg/Na (Table 2). This factor can be interpreted as silicate weathering. The samples with high Pt had 87Sr/86Sr ratios similar to the global average river water (0.7111) [Peucker-Ehrenbrink et al., 2010] and between Phanerozoic marine carbonate (0.7067–0.7091) [Denison et al., 1998] and UCC (0.716) [Goldstein and Jacobsen, 1988] (Figure 3). This indicates that the outlier samples are not related to extensive involvement of old basement at this regional scale. Weathering of young volcanics or silicates with UCC composition can explain much of the variance in Pt concentrations observed in the outliers.
 The reported Pt content in potential source lithologies are: UCC (0.510 ng g−1) [Peucker-Ehrenbrink and Jahn, 2001], MORB (0.2–0.5 ng g−1), reducing sediments on continental margins (0.6–3.5 ng g−1), and metalliferous sediments (1.6–35.5 ng g−1) [Ravizza and Pyle, 1997]. Unweathered organic-rich shales can be up to 9 ng g−1 [Peucker-Ehrenbrink and Hannigan, 2000]. Very high values are observed for organic- and sulfide- rich shale (up to 150 ng g−1) [Orth et al., 1988] and Cu-Ni-PGE or chromatite deposits [Maier, 2005; Mungall and Naldrett, 2008; Naldrett et al., 2008]. However, in the absence of strong ligands (e.g., thiosulfate or organics) [Anthony and Williams, 1993; Bowles et al., 1995; Wood, 1996; Wood et al., 1992] solubility in aqueous media is limited. This is also consistent with Coker et al.'s  observation of rapidly decreasing Pt from 14 pM in waters related to the gossanous zones to below their detection limit of 1 pM within a few meters. In future studies, the role of organic ligands should be explored further in natural systems.
 We examined the relationship between Pt concentration and total dissolved cations (TZ+). Most of the data cluster in a narrow area with Pt concentrations less than 1 pM and TZ+ less than 5 mEq (Figure S8a). The high-Pt samples have low-to-moderate TZ+ values, except CJ239, CJ238 and CJ245 draining arid regions of the Tibetan Plateau, CB101 and CB106 Mt. Baekdu spring samples, and UL713 draining the evaporites of the Russian Far East. These rivers have low or undetermined discharge and therefore their effect on the riverine Pt yield is not significant (Figure S8b). The Si/TZ+ ratio is used as a silicate weathering intensity index [e.g., Huh and Edmond, 1999]. There is no correlation between Pt concentration or yield and Si/TZ+ (Figures S8c and S8d). This suggests that the release and transport of Pt is not a simple function of weathering rates or the intensity of silicate weathering.
 Anthropogenic Pt contamination appears to be global, as demand of Pt has increased in the past several decades and its increased deposition has been observed in urban lake and peat bog sediments [Rauch et al., 2004a, 2004b] as well as in remote Antarctic snow samples [Soyol-Erdene et al., 2011a]. Lacking data for other trace elements with potential anthropogenic sources (either from mining or from atmospheric emission) we cannot attempt to resolve the anthropogenic component of the Pt found dissolved in rivers. There may have been a rise in the background level due to aerosol transport, and the possibility exists that the riverine flux includes a significant “contamination” component above the natural level. At least, we do not observe a gross Pt contamination of surface waters in East Asia, perhaps because automobile catalysts have been introduced relatively late into this region and much of the area sampled lacks significant traffic.
3.2. Riverine Pt Flux to the Ocean
 To calculate the dissolved Pt flux of each of the eleven river systems, the median concentrations of each river system were multiplied by their respective mean annual discharges at the mouths [Fei, 1999; Gaillardet et al., 1999; Huh et al., 1998b]. Values for the mouth rather than the furthest downstream sample were used, because for the eastern Tibetan Plateau rivers our furthest downstream samples were significantly upstream of the mouth. The relatively homogeneous median values, barring the 12 outliers none of which are main channel samples, justified the choice of median values. Estimated dissolved fluxes of Pt were between 1 mol y−1 (Duman) and 325 mol y−1 (Chang Jiang), the variability arising mainly from that of water discharge (Table 3). The discharge-weighted mean concentration, calculated based on the median Pt concentrations for the eleven rivers and their discharges at mouth, was 0.36 pM (Table 3). This is similar to the seawater concentration of Pt (0.3–1 pM) [Colodner et al., 1993; Hodge et al., 1985; Jacinto and van den Berg, 1989].
Table 3. Dissolved Pt Fluxes of the Major Rivers of East Asia
|River System||No. of Samples Analyzed||Areaa||Dischargea||Runoff||Pt (median)b||Pt Flux||Pt Yield|
|103 km2||km3 y−1||mm y−1||pM||mol y−1||10−3 mol km−2 y−1|
|Total||198||9438||2873|| || ||1023|| |
|Discharge-weighted mean|| || || || ||0.36|| ||150|
 The dissolved Pt yield, the flux per unit area, was approximately three times higher in the eastern Tibetan Plateau rivers (Salween, Mekong, Chang Jiang, Hong, and Huang He) than in the rivers of the Russian Far East (Amur, Lena, Yana, Indigirka, and Kolyma) (Table 3). The caveat is that the median Pt concentrations of some rivers are based on only a few samples.
 Because dissolved Pt concentrations did not show appreciable variability in our ∼200 samples covering various lithologies and climatic zones of East Asia, we assumed the discharge-weighted mean (0.36 pM) to represent the global riverine influx to the ocean. When multiplied by the global river runoff of 37,400 km3 y−1 [Palmer and Edmond, 1989], the riverine inflow of dissolved Pt is 13 × 103 mol yr−1.
 Total riverine flux for other PGEs are 1.6 × 103 mol Os yr−1 [Levasseur et al., 1999] and 0.4 × 103 mol Ir yr−1 [Anbar et al., 1996]. Based on the riverine fluxes of Pt (this study), Os, and Ir, the molar Pt/Os ratio for the dissolved riverine flux is 8.1 and is about half that of the UCC (Pt/Os = 16) [Peucker-Ehrenbrink and Jahn, 2001] and higher than that in surface seawater (5.2) [Ravizza, 2001]. The Pt/Ir ratio for the dissolved riverine flux is 30, considerably higher than its UCC ratio of 24 [Peucker-Ehrenbrink and Jahn, 2001] but one order of magnitude lower than that in surface seawater (520) [Ravizza, 2001]. The differences in Pt/Os and Pt/Ir ratios between the river and surface seawater are likely related to the different mobility in aqueous environments, precipitation mechanisms in the estuaries and/or in the oceans [Goldberg et al., 1986; Sharma and Wasserburg, 1997], and contributions of possible unknown sources of these elements to the ocean [Peucker-Ehrenbrink, 2002].
3.3. The Oceanic Mass Balance
 The oceanic residence time of Pt (τPtsw) can be estimated from the rate of addition of Pt to seawater as follows (Figure 4):
where CPtsw is the estimated concentration of Pt in seawater (0.26 pM) [Ravizza, 2001], Mo is the mass of the oceans (1.34 × 1021 kg), the F terms are the Pt fluxes into the ocean, friver is the fraction that survives the estuary and reaches the deep ocean, and faeolian and fmeteorite are the fractions that dissolve. The submarine groundwater discharge is becoming recognized as important for many elemental cycles, but data for Pt is as yet unavailable.
 The main continental input is probably through rivers but there is a lack of direct data from the estuaries making it difficult to estimate friver. Considering the results of laboratory particulate-aqueous partitioning experiments [Cobelo-Garcia et al., 2008; Turner, 2007] and the 3-point estuarine transects [Obata et al., 2006], a 50% release (friver = 1.5) yields 20000 mol y−1. However, frivercan be much higher if the river carries high levels of suspended sediment, if the particulate-aqueous distribution coefficient is high, and/or if the riverine Pt concentration is high. On the other hand, Pt could also be taken up in the estuaries through interaction with organic matter. This would be favored in systems with longer residence time in estuaries where the slow kinetics for Pt coordination with organic ligands [Cosden et al., 2003] can be overcome. A 50% uptake of dissolved Pt in the estuaries (friver = 0.5) yields 7000 mol y−1.
 The delivery of Pt to the ocean by aeolian dust was calculated by taking the dust deposition rate to the ocean of 450 Tg y−1 [Jickells et al., 2005], assuming the dust Pt content to be that of the UCC (0.510 ng g−1) [Peucker-Ehrenbrink and Jahn, 2001], yielding 1200 mol y−1. For faeolian, we assumed that 3% of the aerosol-borne Pt dissolves in the water column by analogy with gold [Falkner and Edmond, 1990]. The result is an aeolian dust contribution to dissolved seawater Pt of 36 mol y−1.
 The mantle input of Pt to the ocean through high temperature hydrothermal plumes is difficult to estimate. McKibben et al.  found up to 2.5 nM Pt in the Salton Sea geothermal brines, and theoretical and experimental studies predict equilibrium concentrations over 500 nM in conditions typical of submarine hydrothermal solutions [Gammons et al., 1992; Wood, 1990]. However, as is the case with iron, most Pt probably forms sulfides around the orifice of hydrothermal vents and fine grained oxyhydroxides close to the ridge axes. Therefore, we have considered hydrothermal fluid source flux to be equivalent to the hydrothermal sedimentation sink flux, 1000 mol y−1 (see below).
 Altogether, by far the most important source of dissolved Pt to the ocean is the input through rivers, and if estuarine effect is ignored (friver = 1), the residence time is 24 kyrs. We evaluated the uncertainty in this estimate by assigning relative uncertainties for the major terms: CPtsw (20%), mean annual runoff (12%) [Schlosser and Houser, 2007], and CPtrw (30%) which yielded a 40% uncertainty on the residence time: 24 ± 10 kyrs. However, if we consider the poorly known friver, the residence time could be from 12 kyrs (friver = 2) to 45 kyrs (friver = 0.5) With currently available information, 16 kyrs (friver = 1.5) would be the best estimate.
 The oceanic residence time of Pt can also be estimated from the rate of removal of Pt from seawater (Figure 4). The primary sink for Pt is believed to be burial in hydrogenous sediments. Goldberg et al.  estimated the average sedimentation rate of Pt based on the Pt content of pelagic sediments (2 ng g−1), density (2 g cm−3), and sedimentation rate (1 mm ky−1). We updated their calculation using new data for Pt in oceanic sediments that has become available in the intervening years. We separated the sedimentation flux into three terms: red clay, carbonate, and hydrothermal.
where Ao is the area of the oceans (3.6 × 1018 cm2), F terms are the burial fluxes per unit area and ared clay and acarbonateare the fractions of the deep-seafloor covered by red clays (0.381) and carbonates (0.471) [Berger, 1976]. For the red clay component, we retained the concentration and sedimentation rate estimate of Goldberg et al. , which resulted in 2800 mol y−1. For the carbonate component, we used the 0.926 ng Pt cm−2 kyr−1 determined by Cave et al.  for the NE Atlantic and assumed that 85% was hydrogenous and not lithogenic. The Pt sedimentation rate in carbonate was higher at 6800 mol y−1. The hydrothermal sedimentation flux is estimated as 1000 mol y−1based on sediments accumulating under the Rainbow hydrothermal plume on the Mid-Atlantic Ridge [Cave et al., 2003]. These add up to 10600 mol y−1. The oceanic residence time calculated using the sink terms is 33 kyrs.
 Evaluating the uncertainty on the residence time based on sink terms is even more difficult than for that based on source terms. Admittedly the assessment was based on extrapolation of data from limited localities. The Pt concentrations in marine sediments have large ranges, and the sedimentation rates also vary widely: 3–60 mm ky−1 for calcareous ooze and 0.3–15 mm ky−1 for red clay [Schulz and Zabel, 2006]. Considering these uncertainties, the residence time based on sink terms can only provide an order-of-magnitude estimate.
 Our estimated oceanic residence time (16 ± 6 kyrs) is lower than the earlier assessment of 106 years using the sink flux [Goldberg et al., 1986]. The residence time of Ir was estimated from a small pristine river in Sweden, the Baltic Sea estuarine environment, seawater concentration based on the Pacific and the North Sea, and burial in sediments, which yielded 2–20 kyrs [Anbar et al., 1996]. Applying the Os/Ir ratio of river water, Sharma re-evaluated the residence time of Ir as 9–14 kyrs. The oceanic residence time of Os was evaluated using the oceanic Os isotope budget as 24–54 kyrs [Levasseur et al., 1999].