Ocean mixing and ice-sheet control of seawater 234U/238U during the last deglaciation

See allHide authors and affiliations

Science  04 Nov 2016:
Vol. 354, Issue 6312, pp. 626-629
DOI: 10.1126/science.aag1015

Uranium in the deep sea

The ratio of 234U to 238U in seawater underlies modern marine uranium-thorium geochronology, but it is difficult to establish the ratio precisely. Chen et al. report two 234U/238U records derived from deep-sea corals (see the Perspective by Yokoyama and Esat). The records reveal a number of important similarities to and differences from existing records of the past 30,000 years. Higher values during the most recent 10,000 years than during earlier glaciated conditions may reflect enhanced subglacial melting during deglaciation.

Science, this issue p. 626; see also p. 550


Seawater 234U/238U provides global-scale information about continental weathering and is vital for marine uranium-series geochronology. Existing evidence supports an increase in 234U/238U since the last glacial period, but the timing and amplitude of its variability has been poorly constrained. Here we report two seawater 234U/238U records based on well-preserved deep-sea corals from the low-latitude Atlantic and Pacific Oceans. The Atlantic 234U/238U started to increase before major sea-level rise and overshot the modern value by 3 per mil during the early deglaciation. Deglacial 234U/238U in the Pacific converged with that in the Atlantic after the abrupt resumption of Atlantic meridional overturning. We suggest that ocean mixing and early deglacial release of excess 234U from enhanced subglacial melting of the Northern Hemisphere ice sheets have driven the observed 234U/238U evolution.

The last deglaciation [18.0 to 10.5 thousand years ago (ka)] saw massive changes in Earth’s surficial environments, including in temperature and precipitation, as well as the retreat of the Northern Hemisphere (NH) ice sheets and sea-level rise (1, 2). These processes have the potential to induce large variability in the weathering of the continents and the composition of chemical fluxes to the ocean. The ratio 234U/238U is one of the isotopic signatures with the potential to have recorded global changes in continental weathering during this critical climate transition (3).

The activity ratio of 234U to 238U in the modern seawater is ~15% higher than secular equilibrium (4) because of the relatively mobile nature of 234U induced by α-recoil effects (5) in the weathered host rocks. 234U is enriched relative to 238U at particle boundaries or damaged lattices and is expected to be preferentially released in the initial phases of weathering (6). Leaching experiments (7) support that the early fraction of granite leachates is high in both U concentration and δ234U [δ234U = (234U/238Uactivity ratio – 1) × 1000)]. In the last glacial period, subglacial drainage of meltwaters from a limited area of ice-sheet interior to the margins may have been possible (8), analogous to the Antarctic Ice Sheet today, where basal meltwater is routed to the margins via subglacial channels (9). Nevertheless, a large fraction of the glaciated NH continents likely had very limited chemical weathering flux to the ocean because of widespread freezing conditions at the ice-sheet base (10). It is reasonable to assume that a labile pool of excess 234U due to α-recoil would have accumulated in the frozen sediments or isolated subglacial lakes and ponds in the wet-based zones under these ice sheets; in fact, high dissolved δ234U of up to ~4000 per mil (‰) has been observed in the Antarctic Taylor Valley (11), a region thought to be hydrologically connected to the nearby ice sheets (12). In the nonglaciated regions, δ234U released by weathering would also respond to tectonic and precipitation changes (3). A reliable reconstruction of oceanic δ234U thus offers a potentially important means for tracing global-scale weathering variability during climate transitions.

The long residence time of U in seawater [~400 thousand years (ky) (13, 14)] would lead to the expectation that any equilibrium response to external inputs should be more than an order of magnitude longer than the deglacial time scale (~10 ky). However, a growing number of studies have indicated that seawater δ234U might have been lower during the last glacial period than during the Holocene and previous interglacials (3, 1517) and might have also changed on millennial time scales (18). These observations imply that there have been large, relatively rapid changes in the U isotope budget of the ocean and are supported by an updated compilation of published initial δ234U data from corals for the last 30 ka (Fig. 1A). However, the extensive scatter in the data, which is likely due to diagenesis (19), has limited the ability to constrain the timing and magnitude of δ234U variations through time.

Fig. 1 Seawater δ234U evolution over the past 30 ky, reconstructed from corals.

(A) Compilation of published coral initial δ234U (δ234Ui) within the range of 135 to 155‰ and with 2σ errors < 3‰ (19). Data outside of this range have been truncated. (B) Low-latitude North Atlantic records with ±2σ confidence lines (black; the green line shows the mean). Also shown are the initial δ234U values for the low-latitude Pacific, reconstructed from Galápagos deep-sea corals (the gold line shows the mean, with the dotted section highlighting the data gap in HS1). One Galápagos coral data point from the B-A with an initial δ234U higher than 155‰ has likely experienced diagenesis and is not shown (19). Black dashed lines mark the modern seawater signature.

To put more robust limits on seawater δ234U evolution and infer past changes in chemical weathering, we report two well-constrained U isotope records for the past 30 ka based on well-preserved deep-sea corals recovered from the low-latitude North Atlantic and the Pacific Galápagos platform (Fig. 1B and figs. S1 and S2), with additional samples for reference from 50 to 30 ka (19) (fig. S3). The general trend of δ234U evolution in these records agrees with previous studies (3, 1517), although the newly sampled glacial-period deep-water corals tend to exhibit higher initial δ234U than the surface corals in each ocean basin (fig. S3). Our data thus suggest a somewhat smaller glacial-Holocene δ234U difference of only 3 to 4‰ (Fig. 1). Atlantic δ234U started to increase around 22 to 20 ka (Fig. 1B), followed by a rapid increase of ~6‰ up to 150‰ during Heinrich Stadial 1 (HS1, 18.0 to 14.6 ka). Available Pacific data are less well resolved, but the δ234U values are lower than those in the Atlantic during early HS1. The δ234U in the Atlantic and Pacific records converged to the modern level during the Bølling-Allerod (B-A, 14.6 to 12.9 ka). Before the last glacial maximum (25 to 50 ka), our data set is consistent with a lower δ234U than that of the Holocene (fig. S3), but it is not well enough resolved to identify potential millennial-scale changes (18).

There are several possible causes for the observed deglacial increase in seawater δ234U. Coastal regions have been inferred to retain U with high δ234U during sea-level highstands (16). Re-dissolution of coastal U has been hypothesized to drive increases in δ234U when sea level rises (16). If that is the case, an increase in seawater δ234U should closely follow sea-level rise. The increase in Atlantic δ234U does appear to have coincided with the initiation of sea-level rise (Fig. 2), but most of the increase in δ234U occurred before the major sea-level rise in the late deglaciation. Hence, we suggest that although U stored in coastal areas could be important for δ234U variability in other situations (18), it was probably not the main driver of the last deglacial δ234U evolution. An increase in deep-water oxygenation may release redox-sensitive U from the seafloor sediments, possibly affecting the seawater U budget during deglaciation. The early deglaciation is thought to have had lower oxygen concentrations than those of the modern day in the upper ocean (<1.5 km) (20). Increased bottom-water oxygenation in the Atlantic and deep Pacific only occurred after HS1 (20) with the resumption of North Atlantic deep overturning (21), which occurred later than the observed increase in δ234U. Together, these results suggest that external sources of excess 234U or ocean mixing have to be involved to explain the observed δ234U variability.

Fig. 2 Seawater δ234U evolution compared with other climate records.

(A) Modeled retreat rate of the Northern Hemisphere (NH) and Southern Hemisphere (SH) ice sheets (33). (B) Modeled basal melting rate of the North American ice sheets (28) and ice discharge of the Laurentide Ice Sheet (29). (C) 231Pa/230Th from North Atlantic deep sediment cores (21, 23). (D) Sea-level history (34). (E) Our reconstructed δ234U evolution in the upper Atlantic and Pacific. Green solid line, Atlantic mean; green dashed line, Atlantic 2σ error; gold line, Pacific mean (dotting, HS1 data gap); gold shading, estimated range of Pacific δ234U. black dotted line, modern seawater δ234U; Gt, gigaton; y, year.

Models (22) and proxies such as 231Pa/230Th (21, 23) support a reduced deep Atlantic meridional overturning circulation (AMOC) (Fig. 2), as well as a reduced surface Gulf Stream (24) and Agulhas leakage (25), during HS1. These processes likely resulted in an upper Atlantic that was more isolated from the rest of the ocean than it is today (19). Increased isolation of the upper Atlantic (~2.0 km, including the depth range of the deep-sea corals in this study) would act to reduce the effective U residence time and allow its δ234U to change more rapidly than in the Pacific. We applied a two-box model, consisting of an upper Atlantic-Arctic box and a “rest of the ocean” box, to study the influence of changing external sources and changing ocean circulation on seawater δ234U (19). Modern high-latitude riverine inputs have considerably higher δ234U than middle- to low-latitude inputs and mainly supply the Arctic and polar North Atlantic (table S1) (19). If the δ234U and U fluxes of all external sources are kept constant throughout the past 25 ky, our model shows that a slowdown of ocean circulation during HS1 can result in a resolvable difference in δ234U between the upper Atlantic and the rest of the ocean, depending on the degree of reduction in exchange flux (Fig. 3A). This result is consistent with the difference in δ234U between the Pacific and Atlantic records (Fig. 1) and contrasts with the general assumption that U isotopes are homogenous throughout the global ocean [i.e., differences no larger than 0.4‰ (4)]. Our result also raises the possibility that other isotope systems with relatively long residence times might exhibit differences between different ocean basins during periods of reduced ocean mixing.

Fig. 3 Sensitivity experiment assessing the response of seawater δ234U to ocean circulation and external inputs.

(A) Effect of ocean mixing slowdown alone (sv, sverdrup). (B) Effect of variable δ234U in inputs to the upper Atlantic, applying a 50% reduction of the glacial exchange flux during HS1. The area between the dashed-dotted curves is the range of low-latitude Atlantic δ234U. External U fluxes through time in both experiments are kept at the modern inputs. The spectrum of colors denotes different experiments (solid lines, upper Atlantic; dashed lines, other ocean). Further details are in (19).

Ocean circulation, however, cannot account for the overall glacial-interglacial increase of ~3‰ or more in the δ234U of both ocean basins (Fig. 3A). External 234U inputs must have increased relative to 238U during the early deglaciation. With three times the modern riverine U flux, a 3‰ shift in oceanic δ234U is possible during HS1 (fig. S4). However, given that the hydrological cycle was probably weaker during the late last glacial period and early deglaciation than at present (2), increased dissolved U flux from global rivers alone as a driver for the δ234U change is unlikely. An increase in δ234U in the continental inputs is therefore likely to have been important in seawater δ234U evolution. There are no direct measurements of the δ234U of past watersheds, so we compiled δ234U data from speleothems that grew during the relevant time period. Although speleothem δ234U does not necessarily reflect the primary surface-weathering signal, because it includes influences such as the percolation of meteoric waters from the surface down into the cave, it may still provide a first-order indication of the variability in the hydrological cycle that controlled the overall 234U budget available for weathering (26). The data come from the middle to low latitudes (fig. S5), and they do not exhibit any distinctive δ234U shifts from 30 to 15 ka, suggesting that weathering variability in these areas was not large enough to account for the oceanic δ234U observations.

Instead, high-latitude processes could have played a key role in driving the increase in global seawater δ234U from the last glacial period to 15 ka. In the early part of the last deglaciation, the bases of North American ice sheets are thought to have become increasingly wet (10). We hypothesize that 234U-enriched water was released from subglacial melt reservoirs with a prolonged residence time and from leaching of the previously frozen subglacial sediments by basal meltwater over this period. In both cases, subglacial meltwaters are likely to be enriched in recoiled 234U (7, 11). In addition, sediments (and their former porewaters) frozen within the bases of icebergs might also have contributed to releasing U and nutrients to the ocean (27) during this period. The reconstructed number of ice streams based on field evidence (8) and the modeled subglacial melt rate (10, 28) (Fig. 2B) of the North American ice sheets were much higher during the early deglaciation (before 15 ka) than later. It is notable that the modeled ice discharge (29) of the Laurentide Ice Sheet was also high during the early deglaciation, reaching a peak at late HS1 (Fig. 2). This timing is consistent with the enhanced release and transport of excess 234U to the ocean during the early deglaciation. In this case, the peak oceanic δ234U indicates that basal water from the whole ice-sheet interior may have been active in exporting to the margin during HS1. In another modeling experiment (19), the sensitivity of oceanic responses to inputs with different δ234U ratios was tested. With an average δ234U of ~800‰ for the input to the upper Atlantic during HS1, the model is able to reproduce the observed amplitude of the increase in δ234U from the last glacial period to 15 ka (Fig. 3B). A more realistic case might be an increase in both the U flux and the δ234U of the high-latitude continental inputs, but deconvolving these two factors is difficult. There is a hint from the compiled data that the surface ocean had lower δ234U than the intermediate-depth ocean (fig. S3). Our extended modeling experiments (table S2) are unable to replicate this feature, even with all high-latitude excess 234U routed to the intermediate ocean via the polar surface Atlantic and no deep-ocean overturning and mixing (19). This indicates that other mechanisms, such as local influences or diagenetic processes, are responsible for those differences.

The transition from HS1 to the B-A is marked by a distinct decrease in δ234U in the low-latitude Atlantic by ~3‰ from about 150‰ to the modern seawater signature. The Pacific δ234U appears to have risen at the same time, converging with the Atlantic δ234U during the mid–B-A. We suggest that this convergence is due to the abrupt increase in overturning of the Atlantic at the end of HS1 (21, 23) (Figs. 2C and 3A), although the depletion of the subglacial excess 234U pool might also have played a role (19). Stabilized Holocene seawater δ234U afterward implies that the U cycle in the modern ocean is likely close to a steady state (30). Nevertheless, perturbation of oceanic U isotopes by polar processes might have continued to occur even during the climatically and oceanographically more stable late Holocene. For example, widespread collapse of the Ross Ice Shelf and the export of old materials from inland Antarctica is inferred to have occurred at ~5 to 1.5 ka in response to regional warming (31). These processes may have been accompanied by a 234U-rich flux to the Southern Ocean, although any shift in the whole-ocean δ234U was probably limited. On longer orbital time scales, seawater δ234U is thought to be higher during past interglacial periods and lower during glacial periods [e.g., (3, 16, 17)], with the decreases potentially associated with low sea level (16). By comparison with the deglacial mechanism, our study implies that progressive NH glaciation could have reduced the weathering input from high-latitude continental areas, leading to lower glacial oceanic δ234U.

The retreat of the NH ice sheets started about 20 ka and continued through HS1 (Fig. 2A) (1), considerably earlier than the enhanced surface melting that dominated ice-sheet mass loss and sea-level rise in the late deglaciation and early Holocene (29). Our data provide evidence for enhanced subglacial melting of the NH ice-sheet interior during the early deglaciation, supporting the notion that basal melting and sliding represents one of the feedbacks involved in enhancing early deglaciation as a result of the buildup of very large NH ice sheets (10). An interesting consequence of the basal melt inputs may be the associated release of nutrients to the ocean. Recent work from the Greenland Ice Sheet indicates that dissolved phosphorus yields are at least equal to those from the Mississippi or Amazon rivers (32). In this regard, nutrients from direct subglacial weathering should be considered in future research as a potential source fueling productivity in the North Atlantic during HS1.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Tables S1 to S8

References (3590)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
Acknowledgments: Data related to this work are available in the supplementary materials. Insightful comments from two anonymous reviewers helped us improve the manuscript. We thank C. Taylor and C. Coath for help in the laboratory. This study was funded by the European Research Council, the Natural Environment Research Council, the Philip Leverhulme Trust, the U.S. National Science Foundation, a Marie Curie Reintegration Grant, and the NOAA (National Oceanic and Atmospheric Administration) Ocean Exploration Trust. We also thank the Charles Darwin Foundation, Galápagos National Park, and INOCAR (Instituto Oceanográfico de la Armada) for making the Galápagos sampling possible.
View Abstract

Stay Connected to Science

Navigate This Article