The ability to reconstruct past ocean currents is essential for determining ocean circulation’s role in global heat transport and climate change. Our understanding of the relationship between circulation and climate in the past allows us to predict the impact of future climate-driven circulation changes. One proposed tracer of past ocean circulation is the neodymium isotope composition (εNd) of ancient water masses. However, ambiguities in what governs the εNd distribution in the modern ocean hamper interpretations of this tracer. Here we present εNd values for marine pore fluids, sediments, and the overlying water column for three sites in the North Pacific. We find that ocean bottom water εNd (εNdBW) in the northeast Pacific lies between the value expected for the water mass (–3.3) and the measured εNd of sediment pore fluid (εNdPW;–1.8). Moreover, εNdPW resembles the εNd of the sediment. Combined, these findings are consistent with recent assessments that sediment pore fluids may be a major source of rare earth elements to the ocean and suggest that the benthic flux of Nd from pore fluids exerts the primary control over the deep ocean distribution of εNd.
Neodymium isotopes are used as a tool in reconstructing ocean circulation. The Nd isotopic value (εNd) is defined as [(143Ndsample/144NdSample) / (143NdCHUR/144NdCHUR) – 1] × 104, where CHUR is the chondritic uniform reservoir, used as an average Earth value (143Nd/144Nd = 0.512638) (Jacobsen and Wasserburg, 1980). Utilizing εNd to reconstruct ocean circulation is based on the fundamental assumption that changes in εNd reflect conservative mixing of water masses (Frank, 2002). However, this assumption has been called into question because the marine budget for εNd is unbalanced (Tachikawa et al., 2003; van de Flierdt et al., 2004; Arsouze et al., 2009) and because water mass εNd (εNdWM) appears to be altered by non-conservative processes in marginal settings or “boundary exchange” (Lacan and Jeandel, 2005; Carter et al., 2012; Grasse et al., 2012; Grenier et al., 2013; Haley et al., 2014; Stichel et al., 2015). In addition, global ocean circulation models incorporating εNd distributions suggest that there is a “missing” source of dissolved Nd that contributes up to ∼95% of the Nd in the ocean (Arsouze et al., 2009). Pore fluid concentration profiles indicate that this missing source could be a benthic flux of Nd from sedimentary pore fluids (Sholkovitz et al., 1989; Haley et al., 2004; Abbott et al., 2015). If so, how does this benthic flux of Nd impact the distribution of εNd in the ocean and the use of εNd in paleoclimate reconstructions? To answer this question, we examine the εNd of the pore fluids, which represent a benthic source of Nd that is significant to the marine Nd budget (Abbott et al., 2015).
We collected 20 L water column, ∼1 L sediment pore fluid, and sediment samples from three sites off the Oregon margin (northwest United States) in October 2012 and July 2013 (Fig. 1; detailed description is in Abbott et al., 2015). Briefly, water column samples were collected using Standard PVC Niskin bottles that were pressurized with N2 upon recovery to filter the sample using inline “Disposal A” 0.45 μm filters (Geotech Environmental Equipment item 73050004). All water column samples were acidified to pH ≤ 2.5 using ultrapure 12 M HCl. Pore fluid was collected using centrifuged sediments from multiple cores, and the centrifuged sediments were then digested in a mixture of HNO3, HCl, and HF using a CEM Corporation MARS 5 microwave (Muratli et al., 2012; Abbott et al., 2015).
Water and sediment digests were analyzed for Nd concentrations on the Thermo VG ExCell quadropole inductively coupled plasma–mass spectrometer (ICP-MS) at the W.M. Keck Collaboratory for Plasma Spectrometry (Oregon State University, USA). A large-volume seawater sample (NBP95R10) collected from Bransfield Strait in the Southern Ocean (62°46′S, 59°24′W) at a water depth of 1300 m was used as an in-house consistency standard (mean 24.8 pM Nd, 1σ = 4 pM, procedural blank 3.5 pM Nd) as no calibrated seawater standards are available. All isotopic analyses were performed on the Nu Plasma ICP-MS multicollector in the Keck Collaboratory at Oregon State University with 144Nd on the axial cup internally normalized for mass bias to 146Nd/ 144Nd = 0.7219. The JNdi-1 standard was used for normalization to 143Nd/144Nd = 0.512115 with a 2σ uncertainty of ±0.000011, n = 166 (reference value 0.512115 ± 0.000007; Tanaka et al., 2000). Specpure, a Nd element quadropole standard, was used as an in-house reproducibility standard with 143Nd/144Nd = 0.511205 with a 2σ uncertainty of ±0.000014, n = 147.
RESULTS AND DISCUSSION
We find water column Nd concentrations to range between 10 pM and 40 pM (Fig. 2A; see the GSA Data Repository1) and εNd to range between −1.2 and −3.2 (Fig. 2B; see the Data Repository). Using potential temperature and salinity data, we interpret our sites at 1200 m and 3000 m depth to represent mixing between North Pacific Intermediate Water (NPIW, ∼240 m depth) and Pacific Deep Water (PDW, ∼3000 m depth; Fig. 2B). Based on these water mass identifications and the published εNd of NPIW and PDW (Haley et al., 2014), the water column εNd profile is predicted to decrease with depth from −3 toward −3.5 (Fig. 2B). Instead, the observed εNd appears to remain constant with depth below the surface, at −2.5 at 1200 m and −2.3 at 3000 m (Fig. 3). The deviation between the observed and the expected ocean bottom water εNd (εNdBW) is greatest (ΔεNdexp–obs = 1.0) in bottom water at our 3000 m site, coinciding with the largest benthic flux of Nd to the ocean (Fig. 3; Abbott et al., 2015). The εNdobs deviates toward the εNd of pore fluids (εNdPW), with εNdobs being less radiogenic than predicted (Fig. 3). The average εNdPW at each site is nearly constant down core and is −0.2 at 200 m, −1.5 at 1200 m (excluding 1.2 cm and 2.4 cm), and −1.8 at 3000 m (Fig. 3). These values are offset from PDW values (−3.5), and instead must be generated from the bulk sedimentary solid phase (Fig. 3; see the Data Repository). Regardless of the mechanism of generation, our measured εNdPW demonstrate that pore fluids can produce an isotopically distinct flux term.
Our model provides a mechanism to explain deep-water εNd alteration in the North Pacific in the absence of modern deep-water formation. Specifically, the benthic flux of Nd from sedimentary pore fluids can alter bottom water εNd to resemble a local sedimentary εNd signature. This alteration of bottom water εNd is most noticeable in regions with distinct weathering provenances. For instance, the benthic flux of Nd from the Amazon River depocenter may explain the resemblance of deep Caribbean water εNd (εNd ≈ –9.2) to the εNd of the Amazon River (εNd ≈ –9; Osborne et al., 2014) instead of the εNd of the Antarctic Intermediate Water (AAIW) (εNd ≈ –11). In general, biogenic phases will have an εNd that resembles surface water (Akagi et al., 2014), meaning ΔεNdFlux–WM in regions of dominantly biogenic sediments will not alter εNdBW in homogenous water columns, such as the Southern Ocean (Stichel et al., 2012), but may alter εNdBW in regions where εNdSurface is different than εNdBW, such as in the North Pacific (Akagi et al., 2014). Moreover, low concentrations of Nd in calcareous sediments minimize the ability of these sediments to change εNdWM (Parekh et al., 1977; Elderfield et al., 1981; Shaw and Wasserburg, 1985). The variation in FNd and ΔεNdFlux–WM (Fig. 3) within our small study region highlights the need for further investigation of the reactions that govern the isotope composition and magnitude of the benthic flux throughout the global ocean.
Our benthic source model is consistent with Nd observations in the modern ocean. For instance, large vertical Nd concentration gradients in the Pacific (Lacan et al., 2012) are consistent with long-term exposure of old bottom water to a substantial benthic flux. Furthermore, the addition of Nd with a pore fluid εNd signature to bottom water would shift the εNd of the bottom water toward more radiogenic values, consistent with deviation of observed bottom water εNd from the predicted bottom water εNd at our sites (Fig. 3). To produce this deviation only requires a small proportion of the dissolved Nd in pore fluids to be transferred to the overlying water column (see Equation 2 below). In contrast, producing the observed bottom water Nd concentrations and εNd with only vertical water column processes is implausible because of the high mass transfer that would be required to create the observed, as much as one unit, shift in εNd, consistent with previous findings (Lacan and Jeandel, 2005; Arsouze et al., 2009; Carter et al., 2012; Grasse et al., 2012; Grenier et al., 2013).
Our model shifts the dominant oceanic source of Nd from a series of surface point sources (i.e., rivers) to a diffuse sedimentary source. Is this hypothesis consistent with the temporal and spatial scales of εNd variability seen in the ocean? To address this question, we use the concept of piston velocity to reflect the ability of the flux to influence the overlying water column. Piston velocity is typically used with respect to gases, but more broadly, piston velocity is a calculation that treats the flux as proportional to the contrast in concentrations (Kump et al., 2005). Here we define piston velocity as the ratio of the magnitude of the flux and the Nd concentration in the overlying water mass (FNd/[Nd]BW). The calculated piston velocities range from 860 cm yr−1 at 3000 m to 360 cm yr−1 at 1200 m to 140 cm yr−1 at 200 m.
The piston velocity should be related to the observed offset between εNdBW and εNdFlux if our calculated piston velocities are indicative of the ability of the benthic flux to alter εNdBW. Accordingly, we find an inverse relationship between the magnitude of the difference between the measured εNdFlux and εNdBW (ΔεNd) and the piston velocity=at our sites. Specifically:
We can go one step further and apply the piston velocity over the height of the overlying water mass to calculate the time required for the benthic flux to change the εNdWM by calculating the “response time” (τresponse). We calculate the response time of a given water mass to the benthic flux of Nd for our model:
Our εNd data and benthic flux model support εNd as a useful tracer of ocean circulation. However, our model adds complexity to interpretations of the εNd data. We propose that εNd retains a memory of its flow path (Equation 1), as seen in our data by the alteration of more radiogenic PDW with a less radiogenic benthic flux (Fig. 3). This memory means that past changes in circulation, such as shifts in the location of deep-water formation and current velocity, are recorded by εNd in part because they change the exposure time to benthic fluxes. For example, sluggish deep-water formation may alter εNdBW simply due to increased exposure time of bottom water to the benthic flux. Additionally, factors such as sediment distribution and composition changes (Scher and Martin, 2004; Fagel and Hilaire-Marcel, 2006; Franzese et al., 2006) may alter εNdBW with no concurrent change in circulation. We suggest that εNd is a more precise tracer of circulation because these factors can be determined and provide additional constraints for reconstructions.
In conclusion, we suggest that marine sediments are a dominant source of Nd to the ocean and that this benthic source likely controls the Pacific Ocean’s deep-water εNd distribution. We assert that the εNd signature of a water mass is determined by a combination of the circulation path and the exposure time to sedimentary fluxes, advancing our ability to interpret εNd as a robust tracer of circulation. We suggest that the influence of boundary exchange through a large benthic flux may continue to alter the εNd signature of a water mass in the deep sea. This model presents a revised way to examine the oceanic Nd budget and suggests that future εNd studies are needed to test this model in the abyssal Pacific and other major ocean basins.
Portions of this work were funded through National Science Foundation grant OCE-1147407 to McManus and Haley. The University of Akron also supported McManus’ contributions to this study. We would like to thank the captains and crew of the R/V Oceanus for their support during the expeditions. We also appreciate many other individuals for their assistance: Clare Reimers, Chris Moser, Jesse Muratli, Paul Walczak, Chris Holm, Meghan Megowan, and Renee Renn helped out during the field and/or laboratory portions of this research; Geomatics Research provided software development to aid in data processing; and Andy Ungerer provided ICP-MS facility support at Oregon State University’s W.M. Keck Collaboratory. Editor Ellen Thomas and three anonymous reviewers provided thoughtful and constructive comments that improved the text.