To investigate the mechanisms of the offset of seawater radiogenic Nd-Hf isotope compositions from those of the upper continental crust rocks, combined Nd-Hf isotope compositions of desert and loess samples from northern China (which integrate a wide range of lithologies and ages of continental rocks) are presented in this study. The results show significant and systematic fractionation of Hf isotopes between fine-grained detritals/leachates (<5 µm) and coarser fractions (<75 µm) of the same samples. A small but systematic difference of Nd isotope compositions between leachates and detrital silicates is also revealed. Overall, the leaching data either plot along or slightly above the Nd-Hf seawater array, providing strong direct support that the seawater Nd-Hf isotope relationship is predominantly generated by weathering of upper continental crust. Our study supports the application of dissolved Hf isotopes as a proxy for different modes of weathering regimes rather than for continental source provenances.
 Since the finding of a strong Hf-rare earth element (REE) fractionation in the Earth's sedimentary system in the 1980s [Patchett et al., 1984; White et al., 1986], a number of studies have contributed to the understanding of Hf isotope geochemistry during sediment transport and oceanic cycling [cf. van de Flierdt et al., 2007; Vervoort et al., 2011]. With more radiogenic Hf isotopes for a given Nd isotope composition, seawater and ferromanganese crusts show a unique correlation (seawater array, Figure 1), which is oblique to the terrestrial array in εNd-εHf space [Albarède et al., 1998]. Despite earlier studies proposing that contributions of radiogenic Hf from hydrothermal sources may be important for the seawater budget of Hf [White et al., 1986; Godfrey et al., 1997; Bau and Koschinsky, 2006], recent dissolved seawater Hf isotope and concentration data have not favored such a scenario [Rickli et al., 2009; Firdaus et al., 2011; Stichel et al., 2012]. However, an unambiguous conclusion has not been achieved due to the fact that hydrothermal contributions of Hf have so far not been measured, and also because there are still no isotopic measurements of truly dissolved versus colloidal Hf isotopes in seawater [e.g., Bau and Koschinsky, 2006]. In addition, Hf isotope data on rivers that directly supply the world's oceans are very scarce. Available data on rivers such as the Hudson (εHf > +30 [Godfrey et al., 2009]) and Moselle basin rivers (εHf < −3 [Bayon et al., 2006]) are actually not consistent with the seawater array (Figure 1), whereas a recent study of four Swiss rivers [Rickli et al., 2013] suggested that the Hf-Nd isotope compositions of the rivers are broadly consistent with the seawater array, but also that the lithologies and even runoff may exert a large influence. It is still possible, however, that the drainage areas of all these rivers are too small to realistically represent the mean upper continental crust (UCC) trend. In this study, we aim to obtain a broader picture of the UCC weathering signal and clarify whether UCC weathering can produce combined Hf-Nd isotope signatures that are consistent with the seawater array.
 The Chinese continent was formed through multiple stages of collision and aggregation of many continental blocks [e.g., Ren, 1996]. Thus, it provides an opportunity to constrain the averaged Nd-Hf isotope signatures of the exchangeable component of upper continental crustal rocks. The dust samples of our study recovered from deserts and other arid areas that distributed over an area extending 600 km from north to south and 4000 km from east to west in northern China are ideal for investigating continental Hf isotope fractionation. The extent of the Chinese deserts and arid areas have been tectonically controlled by the major orogenic belts in East Asia and have been associated with Asian aridification over the late Cenozoic [Chen et al., 2007; Chen and Li, 2011]. Despite that the exact source areas of Chinese deserts and loess deposits are still a matter of debate, these eolian materials always integrate a large range of lithologies (including recycled sedimentary rocks) exposed at the Earth's surface. Dust particles are commonly coated with Fe-Mn oxyhydroxides (the so-called desert varnish [e.g., Bayon et al., 2004]) and trace metal evidence indicated that the desert varnish components are largely formed through natural aqueous leaching of the dust materials [Thiagarajan and Lee, 2004]. Therefore, leaching experiments conducted on these dust samples are assumed to be representative of the Nd-Hf isotopic composition of the labile component of UCC rocks ultimately released by chemical weathering, which provides valuable information on the formation mechanism of the seawater Nd-Hf isotopic array.
2 Materials and Methods
 The desert and loess samples of this study were collected through the efforts of the Institute of Surficial Geochemistry at Nanjing University, and have been described in detail elsewhere (Chen et al., , Li et al., ; see Supporting Information Figure S1 and Table S1).
 Before leaching, a decarbonation step was carried out using excess 0.5 M acetic acid for 8 hours. This is expected to remove the detrital and secondary carbonates while leaving the labile Fe-Mn oxyhydroxide fraction of the silicate grains untouched. Then a dilute reducing and complexing solution consisting of 0.005 M hydroxylamine hydrochloride, 1.5% acetic acid, and 0.03 M EDTA-2Na, buffered to pH about 4 with superpure NaOH was used to extract the mobile component of the dust for about 2 h [Chen et al., 2012]. After complete removal of Fe-Mn oxides, the silicate residues, of which paired leachate data (<75 and <5 µm size fractions) have been obtained, were totally dissolved in steel jacketed autoclaves at ~180–200°C for 4 days. Subsequent purification and measurement of the Nd and Hf isotopes followed Chen et al.  (see Table S1 for more details).
 The Hf and Nd isotope data of dust leachates and detrital fractions are shown in Table S1 and Figure 1.
 There is virtually no difference between the Nd isotope composition of coarse- and fine-grained dust leachates (Figure 2a). In addition, the total dissolution data of the fine- and coarse-grained silicate fraction further reveal that their Nd isotope compositions are also distributed very close to the 1:1 line (Figure 2b). Interestingly, a small but detectable and systematic difference between the Nd isotope compositions of the leachates and detrital silicate fractions (about 1.4 epsilon units) is remarkable (Figure 2c).
 With the exception of Taklimakan leachates, which have highly radiogenic Hf isotope signatures (εHf: +14.5, +17.0), most of the εHf values of the leachates are in the range of −0.8 to +9.9, while their corresponding εNd signatures vary from −15.4 to −3.4. The most striking feature of the leaching data is that they all fall along or slightly above the seawater array (Figure 1). Also, the hafnium isotope compositions are more radiogenic in leachates of the coarser size fraction (<75 µm) than the corresponding finer grain size leachates (<5 µm) (Figures 1 and 2d).
 As expected, the detrital silicates of the <75 µm size fraction has Nd and Hf isotope signatures consistent with the crustal array in εHf-εNd space, while the <5 µm fraction is more radiogenic in its Hf isotope composition (Figures 1 and 2e). Moreover, the leachate Hf isotope data of the fine and coarse size groups deviate notably from such a 1:1 line (Figure 2d). Also, the Hf isotope compositions of the leachates are fractionated significantly from those of the silicates (Figure 2f).
4 UCC Weathering Could Fully Produce the Seawater Array
 Interestingly, the leachate of a typical Asian dust sample GPC3 recovered from the North Pacific (Table S1) [Pettke et al., 2002] shows εNd and εHf signatures of −3.8 and +6.6, respectively (Figure 1), which are entirely consistent with central North Pacific deep water compositions [Amakawa et al., 2009]. Thus, the Nd-Hf isotopic signal of the leaching experiments of GPC3 supports that the leaching protocol is robust in extracting the exchangeable component (e.g., Fe-Mn coatings) of the Asian dust (Figure 1).
 Supporting evidence for the agreement of the dust leaching signal with seawater array comes from the eastern equatorial Atlantic, which receives high loads of Saharan dust input. The Nd-Hf isotopes of eastern equatorial Atlantic surface waters (5°N–15°N, Figure 1) were suggested to be dominated by dissolution of Saharan dust [Rickli et al., 2010]. The consistency of leached Nd-Hf isotope trend between the fine-grained dust of this study and the Saharan dust implies that over a wide range of source rock compositions and ages, as well as geographical locations, aquatic Hf isotope fractionation of dust particles can provide Hf isotope signatures radiogenic enough to match the seawater array. Note that the leachates reflect Hf isotope signatures released during weathering of large-scale averaged upper continental crust (in our case represented by the dust materials). Therefore, it is reasonable to assume that Hf isotope signatures released from (large-scale averaged) rivers or shelf sediments are similar to the dust leachates. Our study thus cannot quantitatively distinguish the specific delivery mechanism of Hf-Nd isotopes to the ocean.
 For the samples with εNd signatures below −13, the Hf isotope compositions of the leachates are more radiogenic than the corresponding value of the seawater array (Figure 1). Because the largest difference between the seawater and crustal arrays (least radiogenic Nd isotope compositions, Figure 1) also indicates the largest degree of incongruent weathering, it is not surprising that actual leaching of Hf isotope signatures from older continental rocks shows more scatter. Given that the least radiogenic Hf isotope signatures of seawater have been found in the Labrador Sea [Rickli et al., 2009], which is surrounded by cold and glacial weathering input, it may be that weathering in such an environment is more congruent [van de Flierdt et al., 2002]. It is also likely that the rocks of northern Canada surrounding the Labrador Sea region had an evolution history different from the old cratons in North China, resulting in different mineralogical composition and thus a different degree of aquatic Hf isotope fractionation in these two regions.
 The leachate data clearly and robustly indicate that no “extra” radiogenic Hf is necessary to produce the observed seawater array. Thus, contributions from hydrothermal Hf sources are probably negligible for the global seawater Hf budget.
5 Implications for Nd and Hf Isotopes as Oceanographic and Provenance Tracers
 The exchangeable fractions of continental particles or their preformed coatings either enter seawater via eolian or riverine transport. On the continental shelf, such a preformed leachable component may become the source of “boundary exchange”, a process proposed in previous studies [Lacan and Jeandel, 2005].
 Due to the small Sm/Nd fractionation in different minerals [Bayon et al., 2006], the negligible fractionation of Nd isotopes between coarse and fine grain sizes of the detrital material (Figure 2b) is not surprising. Leaching of the dust particles, in contrast, appears to produce small but systematically more radiogenic Nd isotope signatures than the corresponding silicate dust fractions (Figure 2c) although no fractionation is observed between the fine and corresponding coarse grain size leachates (Figure 2a). Given that stable isotope fractionation cannot cause such a difference and that the good correlation in Figure 2c precludes external sources, our preferred explanation is that this pattern reflects the time integrated effect of slightly higher Sm/Nd in the leachable component during past weathering/sedimentary cycles. For example, the typical Middle Rare Earth Element enrichments during aquatic natural leaching of dust [Thiagarajan and Lee, 2004] would lead to a higher Sm/Nd ratio of the mobile component (e.g., preformed Fe-Mn coatings) than that of the silicate residue. Nevertheless, it is possible that the slightly more radiogenic leaching Nd signal only represents mean state upper continental crust weathering especially in the middle to low latitudes, while recently glacially eroded rocks/till may yield less radiogenic Nd isotope compositions of the leachates than the bulk silicates in a cold climate [Andersson et al., 2001].
 Hafnium isotope fractionation during mineral sorting and leaching is significantly larger than that of Nd isotopes. Overall, the Hf isotope fractionation between leachates and corresponding detrital material is larger the older the source rocks are (i.e., less radiogenic Nd isotopes, Figure 1), reflecting the time integrated effect of Hf isotopic differences between different minerals. The Hf isotope signatures of the <5 µm silicate (Figure 1) are close to the seawater array (though still not radiogenic enough). Interestingly, in marine settings, essentially all detrital sediments fall below the seawater array [Bayon et al., 2009; Vervoort et al., 1999, 2011, Figure 1]. Given that the <5 µm size samples in this study are already close to the lowest size range observable in natural sediments and are also expected to have experienced maximum heavy mineral loss during weathering and transport (including zircon), we propose that the <5 µm size fraction reflects the upper limit of Hf isotope fractionation due to zircon loss. Furthermore, the <5 µm fractions (Figure 1) are close but still systematically more radiogenic in Hf isotope composition than the “zircon-free sediment array” [Bayon et al., 2009]. This implies that the marine fine-grained sediments on average, though clearly depleted in zircons, may not be completely “zircon-free.”
 Surprisingly, Hf isotopes are systematically more radiogenic in the leachates of the coarse grain size fraction than of their fine counterparts (Figure 2b). Given that the leachable component is not necessarily the product of current weathering in the arid deserts, one explanation is that the <5 µm particles may mainly come from a more deeply weathered environment than the large particles (<75 µm). These <5 µm particles would thus have the potential of receiving somewhat higher contributions from zircon weathering. However, such an explanation is unlikely because Nd isotope compositions of the fine and coarse particles are essentially identical. Therefore, our data imply that incongruent weathering of minerals other than zircon are also important for Hf isotope fractionation during aquatic leaching of UCC rocks and consequently in the formation of Nd-Hf seawater array, which is consistent with recent studies [Bayon et al., 2006, 2009; Rickli et al., 2009; Chen et al., 2011]. Unfortunately, the exact contribution of different minerals to the Hf isotope fractionation is unknown and remains to be studied in detail before an unequivocal explanation for this grain size effect on Hf isotopes can be provided.
 Recent direct water mass studies on dissolved Hf isotopes have indicated that Hf isotopes are not a sensitive water mass tracer due to their limited isotopic variability in the oceanic water column [Stichel et al., 2012]. Early studies on Fe-Mn crusts [van de Flierdt et al., 2004] and a recent study on authigenic sediment fractions from the Arctic Ocean [Chen et al., 2012], however, still showed a broad positive correlation between evolution of Hf isotopes and Nd isotopes in seawater, which is also reflected by the well constrained seawater array. As shown in Figure 1, it is evident that there is a large range in the leached Hf isotope compositions from the same rocks which essentially prevents clear identification of source rocks from the dissolved Hf isotope composition of seawater. Consequently dissolved seawater Hf isotope signatures and their records in authigenic sediment fractions will be most useful in distinguishing different modes of weathering regimes on land or even changes in runoff [Rickli et al., 2013].
 We thank J. Heinze (GEOMAR Kiel), U. Westernströer and C.-D. Garbe-Schönberg (Institute of Geosciences Kiel) for their kind help during the experiment. T.-Y. Chen acknowledges the China Scholarship Council for providing fund to his overseas study. G. Li is supported by National Science Foundation of China (41102103, 41021002, and 41173105).