Radioisotope constraints of Arctic deep water export to the North Atlantic

The export of deep water from the Arctic to the Atlantic contributes to the formation of North Atlantic Deep Water, a crucial component of global ocean circulation. Records of protactinium-231 (231Pa) and thorium-230 (230Th) in Arctic sediments can provide a measure of this export, but well-constrained sedimentary budgets of these isotopes have been difficult to achieve in the Arctic Ocean. Previous studies revealed a deficit of 231Pa in central Arctic sediments, implying that some 231Pa is either transported to the margins, where it may be removed in areas of higher particle flux, or exported from the Arctic via deep water advection. Here we investigate this “missing sink” of Arctic 231Pa and find moderately increased 231Pa deposition along Arctic margins. Nonetheless, we determine that most 231Pa missing from the central basin must be lost via advection into the Nordic Seas, requiring deep water advection of 1.1 – 6.4 Sv through Fram Strait.

D eep water export from the Arctic to the North Atlantic through the Nordic Seas affects the formation of North Atlantic Deep Water (NADW) 1,2 , a critical component of thermohaline circulation. Because NADW plays a prominent role in the oceanic transport of latent and sensible heat and the sequestration of atmospheric CO 2 , a well-defined record of Arctic contributions to NADW formation is critically important.
The ratio of 231 Pa (t 1/2 = 3.2 × 10 4 y) and 230 Th (t 1/2 = 7.5 × 10 4 y) in deep-sea sediments can be used as a proxy for deep water circulation and help constrain changes in Arctic outflow. These two naturally occurring radioactive isotopes are produced constantly and ubiquitously in the water column through the decay of uranium isotopes. Due to their long oceanic residence times, uranium isotopes have relatively uniform concentrations and produce 231 Pa and 230 Th in a constant ratio of 0.093. Unlike their soluble uranium parents, 231 Pa and 230 Th are particle reactive, and their contrasting removal rates by adsorption onto settling particles (scavenging) can be exploited to gain information about ocean circulation and particle fluxes.
Thorium-230 has a high particle affinity, and thus a short oceanic residence time of~20-40 years 3 . As a result, this isotope is generally removed from solution in the same geographic location as it is produced, such that the inventory of 230 Th in sediments balances its production in the overlying water column 4 . Protactinium-231 is somewhat more soluble, with an oceanic residence time of~100-200 years, and can therefore be preferentially affected by lateral transport, including advection, before scavenging and deposition in sediments 3 . This differential scavenging results in sedimentary 231 Pa/ 230 Th ratios that diverge from the water column production ratio of 0.093 3,[5][6][7] . In areas of low particle flux, some 231 Pa can escape scavenging and be laterally transported, resulting in low 231 Pa/ 230 Th ratios (<0.093) in underlying sediments. In areas of high particle flux, such as ocean margins, a greater proportion of dissolved 231 Pa and 230 Th are removed via scavenging (a process called boundary scavenging 5,8 ). Because there may be additional 231 Pa in the water column that was transported from areas of low particle flux, this enhanced removal can result in sedimentary 231 Pa/ 230 Th ratios > 0.093.
Sedimentary 231 Pa/ 230 Th ratios across the central Arctic are low [9][10][11][12][13][14][15] , indicating that 231 Pa is laterally transported away from this region of low particle flux. However, the low spatial resolution of sedimentary observations along the margins of the Arctic basin has made it difficult to determine if the 231 Pa missing from the central basin is deposited in margin sediments as a result of boundary scavenging or if 231 Pa is being exported out of the Arctic. The Fram Strait is the only deep conduit in the Arctic, connecting the Eurasian Basin to the North Atlantic through the Nordic Seas (Fig. 1), thus a loss of 231 Pa from the Arctic indicates southward advection of deep water into the Nordic Seas. Finding the "missing sink" of 231 Pa is therefore essential to the interpretation of sedimentary 231 Pa/ 230 Th ratios as a proxy for deep water transport out of the Arctic 10,11 .
Here, we expand the geographic coverage of 231 Pa/ 230 Th measurements in Arctic sediments and find slightly elevated 231

Results and discussion
Coretop 231 Pa/ 230 Th ratios. To improve the Arctic-wide 231 Pa budget, we have expanded the coverage of 231 Pa/ 230 Th measurements in surface sediments along the margins of the Arctic Ocean (continental slopes on the periphery of the Eurasian and Amerasian Basins) and in the central Canada Basin. Because the majority of our samples were collected from elevated locations near continental margins where sedimentation rates are high, we assume that the 231 Pa/ 230 Th ratios are not appreciably influenced by radioactive decay, bioturbation, or turbidites. Although sedimentation rates are significantly lower in the basins, these assumptions are unlikely to change the final 231 Pa/ 230 Th ratios by more than a few percent (see Supplementary Discussion). 231 Pa/ 230 Th ratios in margin sediments are generally higher than those in the basin (Fig. 2). The average 231 Pa/ 230 Th ratio based on previously published data from the central basin 9-15 was 0.063 ± 0.003 (±SE, n = 36); this ratio is not significantly changed by the addition of our new data, increasing slightly to 0.068 ± 0.003 (n = 45) (Fig. 3). However, the addition of 40 new samples at the margins notably increases the average margin ratio from 0.077 ± 0.007 (n = 15; previous studies 10,15,16 ) to 0.101 ± 0.005 (n = 55). Because the surface sediment samples are not evenly distributed throughout the study area, we also determined areanormalized averages by gridding the data into equal-area bins ( Supplementary Fig. 1). This area-normalization does not change the margin 231 Pa/ 230 Th (0.101 ± 0.005; n = 40), and only slightly decreases the basin average to 0.065 ± 0.004 (n = 33).
Protactinium-231 mass balance. The addition of many new margin samples with 231 Pa/ 230 Th ratios near or slightly above the water column production ratio of 0.093 provides evidence of boundary scavenging of 231 Pa: the enhanced accumulation of 231 Pa in sediments resulting from higher particle fluxes. To quantify how much of the 231 Pa missing from the central basin is deposited along the margins, we follow the mass balance approach developed by Luo and Lippold 16 , which first assumes that all of the 231 Pa missing from the central basin is balanced by boundary scavenging along the margins, in order to determine the expected margin 231 Pa/ 230 Th ratio ( 231 Pa/ 230 Th margin ). This expected 231 Pa/ 230 Th margin is then compared to the observed 231 Pa/ 230 Th margin to determine the fraction of 231 Pa that can be explained by boundary scavenging. Importantly, this method assumes that the inventory of 230 Th in sediments is balanced by its inventory in the overlying water column (i.e., any changes in the 231 Pa/ 230 Th ratio are due to the addition or removal of 231 Pa). Previous studies have shown minimal export of 230 Th from the Arctic, supporting this assumption 10,13,17 .
Comparing the area-normalized average 231 Pa/ 230 Th in the basin ( 231 Pa/ 230 Th basin ; 0.065 ± 0.004) to the expected production ratio of 0.093 and assuming that 100% of the 230 Th produced is buried in underlying sediments 10,13,17 indicates that 70 ± 4% of the 231 Pa produced in the basin is buried there. If all of the missing 231 Pa is deposited along the margins, the expected 231 Pa/ 230 Th margin is 0.305 ± 0.024. However, the observed areanormalized average 231 Pa/ 230 Th margin is 0.101 ± 0.005, thus only 4 ± 1% of the 231 Pa produced in the basin can be accounted for at the margins, and the remaining 26 ± 2% must be exported to the Nordic Seas.
Recent water column measurements of 231 Pa support a net export of 231 Pa through Fram Strait 15 , and previous Arctic-wide assessments of surface sediments (based on very few 231 Pa/ 230 Th measurements along the margins) concluded that~30-40% of the 231 Pa produced in the Arctic is exported 10,11 . The greatly improved spatial coverage in this study therefore provides a much better constrained (and slightly decreased) estimate of 231 Pa leaving the Arctic of~26%, but upholds the conclusion that advection through Fram Strait is the main sink for the 231 Pa that is missing from central basin sediments.  Wider sections of the plot indicate a higher probability that a data point in the group will have that value, while narrow sections indicate a low probability that the data in the group falls in that range.  This estimate is within the lower bounds of the significantly larger range of fluxes inferred previously from geostrophy and ADCP analyses, which range from 1.8 to 18.4 Sv 18,28,29,31 , depending on the density or depth range chosen and the measurement approach. Our geochemistry-based estimate also offers an advantage over more direct volume measurements, because the activities of 231 Pa reflect conditions integrated over the water column residence time of this isotope (~100-200 years), unlike snapshot measurements. Our 231 Pa-based estimate could be improved by increasing the number of dissolved 231 Pa measurements in the Greenland Sea, which would refine the estimate of the northward 231 Pa flux through Fram Strait. If the 231 Pa activity of inflowing water is higher than the range estimated here (0.11-0. 16  The revised Arctic 231 Pa budget presented herein provides a geochemical constraint on modern circulation in Fram Strait while also providing evidence that the 231 Pa/ 230 Th ratio can be applied as a proxy for past changes in Arctic circulation. Given these improved constraints on the 231 Pa/ 230 Th ratios of Holocene Arctic sediments, future work should focus on developing downcore records to investigate how changing Arctic outflows may have influenced the formation of NADW through time. In particular, further constraints on the amount of Arctic ventilation during the Last Glacial Maximum are needed to resolve conflicting records of persistent deep water export 11,16 and indications of an isolated, stagnant Arctic basin during this period 42 . Isotopic analyses. Thorium ( 230 Th, 232 Th), protactinium ( 231 Pa), and uranium ( 234 U, 238 U) concentrations were measured using isotope dilution and mass spectrometry on an Element 2 inductively coupled plasma mass spectrometer (ICP-MS) at the Lamont-Doherty Earth Observatory of Columbia University. Approximately 100 mg of sediment was spiked with 229 Th, 233 Pa, and 236 U before digestion in HClO 4 , HNO 3 and HF 43 . Isotopes were separated by co-precipitation with Fe(OH) 3 followed by anion exchange column chemistry (AG1-X8 anion resin) 44,45 . Reproducibility was assessed using internal standards of homogenized Arctic sediments and North Atlantic sediments; relative standard deviations were <6% for all isotopes. Background contamination was corrected for with Milli-Q water blanks that were digested and analyzed using the same procedure as samples.

Methods
This corrects for 230 Th and 231 Pa sourced from continental material rather than produced via U decay in the water column. The lithogenic 230 Th activity was determined for each sample using the activity ratio of 238 U and 232 Th in lithogenic material, ( 238 U/ 232 Th) lith , and the measured activity of 232 Th (dpm/g), which is solely of lithogenic origin 46 : Similarly, lithogenic 231 Pa activity was calculated using Eq. 3 multiplied by 0.046, the natural 235 U/ 238 U activity ratio: 231 Pa lith ¼ 0:046*ð 238 U= 232 ThÞ lith * 232 Th ð4Þ This approach assumes that 238 U and 235 U are in secular equilibrium with 230 Th and 231 Pa, respectively, and that ( 238 U/ 232 Th) lith is known. Previous pan-Arctic studies have used ( 238 U/ 232 Th) lith ratios between 0.5 and 0.7, most commonly 0.6 ± 0.1 10,11,15,16 . However, the Amerasian Basin is likely to have a higher average ( 238 U/ 232 Th) lith than the Eurasian Basin due to the presence of detrital carbonates in the Canadian Shield 47 , which can have 238 U/ 232 Th ratios as high as 4 48 . The bulk sediment 238 U/ 232 Th ratios measured in our samples were higher in the Canada Basin compared to the Makarov, Nansen, and Amundsen Basins, in line with previous studies 10,14 (Supplementary Fig. 2). Due to this geographic variation, we use a 238 U/ 232 Th ratio of 0.7 ± 0.1 to correct for lithogenic inputs in the Canada Basin, and a ratio of 0.6 ± 0.1 to correct for lithogenic inputs elsewhere. Using a higher lithogenic ratio causes the 231 Pa xs / 230 Th xs ratio (referred to hereafter and in the main text as 231 Pa/ 230 Th ratio) to increase slightly.
Authigenic 230 Th and 231 Pa (produced through the in-situ decay of U that has precipitated in reducing sediments) were not accounted for because we assume a surface sediment age of 0 ka, thus no time has elapsed during which 230 Th and 231 Pa could have accumulated from authigenic U decay (see Supplementary Discussion).
Errors on individual 231 Pa/ 230 Th ratios are reported as 2σ. As the errors on the ICP measurements are only a few percent, the largest sources of error are the sizeable lithogenic correction and the error on the detrital ( 238 U/ 232 Th) lith ratio. Errors on averaged 231 Pa/ 230 Th ratios are reported as standard error (SE = σ= ffiffiffi n p ), to reflect increased confidence in the value as data coverage improves.
To compare our results with published data, we recalculated previously reported 231 Pa/ 230 Th ratios in the Canada Basin using a ( 238 U/ 232 Th) lith ratio of 0.7 when possible 10,11,15 . For two samples in the Canada Basin near the Alpha-Mendeleev Ridge, total 231 Pa and 230 Th were not reported and excess activities were based on measured uranium activities rather than the ( 238 U/ 232 Th) lith ratio 12,14 .
Margin-basin designation. Samples were separated into central basin and margin bins by eye based on proximity to the continental slope. As ridges are the second largest physiographic province in the Arctic (preceded by continental shelves) 49 , it is impossible to use a cutoff based on depth or distance to an isobath; this approach would incorrectly place many of the samples collected from ridges into the margin category. Sensitivity tests were performed to ensure that samples located in areas that could be considered part of either bin (e.g., near the Chukchi Plateau, Lomonosov Ridge near Siberia) did not change the conclusions of this study. In all cases, moving these samples from one bin to another changed the average 231 Pa/ 230 Th ratios of the margin and basin bins by <2%. Samples located in the Greenland Sea and on the shallow shelf (<400 m) were not included in either the basin or margin groups (n = 4).
Equal-area grid. Area-normalized averages were determined by gridding the data into equal-area bins (Supplementary Fig. 1). Each bin has a height of 1-degree, and an area of~12,400 km 2 . The longitude range of each bin grows larger toward the pole, as the number of bins per degree of latitude decreases. The assignment of each bin as basin or margin is based on the location of the samples inside the bin, rather than the center of the bin. Changing the bin size does not notably affect the results; doubling the height of each bin from 1 to 2-degrees of latitude increases the average margin 231 Pa/ 230 Th ratio by 5% and does not change the average basin ratio.
In reality, not all of the 231 Pa missing from the central basin is deposited along the margins, but some is exported to the Nordic Seas. To determine the fraction of missing 231 Pa that can be found at the margins (f), the observed margin ratio is compared to the expected value determined in Eq. 5 :16 f ¼ ðmeasured 231 Pa= 230 Th margin À 0:093Þ ðexpected 231 Pa= 230 Th margin À 0:093Þ 100% ð6Þ The volume of the central basin was determined by drawing a polygon following the 1500 m isobath but including the ridges, such that regions shallower than 1500 m inside the polygon were considered part of the basin (Fig. 1). The volume inside the polygon (V basin ) is 1.1474 × 10 7 km 3 (based on the ETOPO2_v2 2 min gridded bathymetry from NOAA; www.ngdc.noaa.gov/mgg/global/etopo2.html). The volume of the margins (V margin ; shelves and slope down to 1500 m) was determined to be 0.1516 × 10 7 km 3 by subtracting the basin volume from the total Arctic volume (1.2990 × 10 7 km 3 ) 50 .
Water column mass balance of protactinium-231. The mass balance of 231 Pa in the Arctic water column is summarized in Eq. 7: Where β 231 is the production rate of 231 Pa (dpm m −3 y −1 ), S 231 is the scavenging rate of 231 Pa (dpm m −3 y −1 ), 231 Pa GS and 231 Pa AO are the activities of 231 Pa in the Greenland Sea and Arctic Ocean, respectively (dpm m −3 ), F GS and F AO are the northward and southward water fluxes through Fram Strait, respectively (m 3 y −1 ), V AO is the total volume of the Arctic Ocean 50 , and ∂ 231 Pa/ ∂t is the change in Arctic 231 Pa concentration through time. At steady state, ∂ 231 Pa/ ∂t is zero. Based on our 231 Pa sediment mass balance, S 231 is equal to 0.74 β 231 . Because both F GS and F AO are unknowns, another equation is needed to solve for the volume exchange through Fram Strait. If we assume that the inflow and outflow of intermediate and deep waters is the same (F = F GS = F AO ), Eq. 7 then simplifies to:

Data availability
The data that support the findings of this study are available through the World Data Service for Paleoclimatology (https://www.ncdc.noaa.gov/paleo-search/study/31932) and in Supplementary Dataset 1.