The Transpolar Drift as a Source of Riverine and Shelf-Derived Trace Elements to the Central Arctic Ocean

Abstract

A major surface circulation feature of the Arctic Ocean is the Transpolar Drift (TPD), a current that transports river-influenced shelf water from the Laptev and East Siberian Seas toward the center of the basin and Fram Strait. In 2015, the international GEOTRACES program included a high-resolution pan-Arctic survey of carbon, nutrients, and a suite of trace elements and isotopes (TEIs). The cruises bisected the TPD at two locations in the central basin, which were defined by maxima in meteoric water and dissolved organic carbon concentrations that spanned 600 km horizontally and ~25–50 m vertically. Dissolved TEIs such as Fe, Co, Ni, Cu, Hg, Nd, and Th, which are generally particle-reactive but can be complexed by organic matter, were observed at concentrations much higher than expected for the open ocean setting. Other trace element concentrations such as Al, V, Ga, and Pb were lower than expected due to scavenging over the productive East Siberian and Laptev shelf seas. Using a combination of radionuclide tracers and ice drift modeling, the transport rate for the core of the TPD was estimated at 0.9 ± 0.4 Sv (10⁶ m³ s⁻¹). This rate was used to derive the mass flux for TEIs that were enriched in the TPD, revealing the importance of lateral transport in supplying materials beneath the ice to the central Arctic Ocean and potentially to the North Atlantic Ocean via Fram Strait. Continued intensification of the Arctic hydrologic cycle and permafrost degradation will likely lead to an increase in the flux of TEIs into the Arctic Ocean.

Key Points

• The Transpolar Drift is a source of shelf- and river-derived elements to the central Arctic Ocean
• The TPD is rich in dissolved organic matter (DOM), which facilitates long-range transport of trace metals that form complexes with DOM
• Margin trace element fluxes may increase with future Arctic warming due to DOM release from permafrost thaw and increasing river discharge

Plain Language Summary

A major feature of the Arctic Ocean circulation is the Transpolar Drift (TPD), a surface current that carries ice and continental shelf-derived materials from Siberia across the North Pole to the North Atlantic Ocean. In 2015, an international team of oceanographers conducted a survey of trace elements in the Arctic Ocean, traversing the TPD. Near the North Pole, they observed much higher concentrations of trace elements in surface waters than in regions on either side of the current. These trace elements originated from land, and their journey across the Arctic Ocean is made possible by chemical reactions with dissolved organic matter that originates mainly in Arctic rivers. This study reveals the importance of rivers and shelf processes combined with strong ocean currents in supplying trace elements to the central Arctic Ocean and onward to the Atlantic. These trace element inputs are expected to increase as a result of permafrost thawing and increased river runoff in the Arctic, which is warming at a rate much faster than anywhere else on Earth. Since many of the trace elements are essential building blocks for ocean life, these processes could lead to significant changes in the marine ecosystems and fisheries of the Arctic Ocean.

1 Introduction

Of all the major oceans on Earth, the Arctic Ocean is the most heavily influenced by processes occurring over continental shelves, which cover over 50% of its area (Jakobsson, 2002). The Arctic Ocean also has the lowest salinity surface waters, a result of limited evaporation, high riverine inputs, the annual sea ice freeze/melt cycle, and restricted exchange with other ocean basins (Serreze et al., 2007). These factors combine to impart a shelf-derived biogeochemical signature over much of the polar mixed layer, the low-salinity surface layer influenced by sea ice and freshwater, even in the central basin.

In the western Arctic's Canada Basin, hydrographic fronts serve as barriers to rapid shelf-basin exchange processes; thereby eddies and wind-induced upwelling or downwelling constitute the primary mechanisms for off-shelf water and material transport and exchange (Muench et al., 2000; Pickart et al., 2005, 2013). In the eastern Arctic, however, the Transpolar Drift (TPD) is a major current that directly transports Eurasian shelf water and sea ice directly from the Laptev and East Siberian Seas toward the central basin and Fram Strait, a major outlet for Arctic waters (Ekwurzel et al., 2001; McLaughlin et al., 1996; Rigor et al., 2002; Rudels, 2015; Schlosser et al., 1994). The time scale for the trans-Arctic crossing of this current is on the order of 1–3 years (Pfirman et al., 1997; Steele et al., 2004); as such, the TPD is currently a mechanism for the rapid transport of shelf-derived materials including nutrients and carbon to the central Arctic Ocean (Kipp et al., 2018; Letscher et al., 2011; Opsahl et al., 1999; Wheeler et al., 1997), with potential biogeochemical impacts detected as far downstream as the North Atlantic Ocean (e.g., Amon et al., 2003; Gerringa et al., 2015; Noble et al., 2017; Torres-Valdés et al., 2013). At present, primary production in the largely ice-covered central Arctic is light limited; however, surface warming has led to reductions in ice cover, as well as increases in river discharge and permafrost thawing (Frey & McClelland, 2009; McClelland et al., 2004; Peterson et al., 2002; Schuur et al., 2015; Spencer et al., 2015). With reduced ice cover, the TPD-derived transport of ice-rafted materials might be interrupted (Krumpen et al., 2019), though Newton et al. (2017) have shown that in the near term (approximately several decades) long-distance ice transport will accelerate as the ice thins and is more responsive to the winds. Together, these changes are expected to modify the ecosystem dynamics of the Arctic Ocean, with shelf-basin exchange processes playing a significant role.

In 2015, three nations led cruises to the Arctic Ocean as part of the international GEOTRACES program, a global survey of the distributions of oceanic trace element and isotopes (TEIs). The Arctic GEOTRACES program represented an unprecedented effort in sampling of the Arctic water column from a biogeochemical perspective. High-resolution coverage of waters above 84°N captured the TEI fingerprint of the TPD and will serve as an important reference for future studies that focus on climate change impacts in the Arctic. Radium isotopes measured during the Arctic GEOTRACES cruises have already been used to show that the chemical composition of the TPD is modified during passage over the Laptev Shelf and to suggest that potentially significant changes in the flux of nutrients and carbon from the Siberian shelves are already underway (Kadko et al., 2019; Kipp et al., 2018; Rutgers van der Loeff et al., 2018). Additionally, Rijkenberg et al. (2018) found higher dissolved Fe and Slagter et al. (2017) found increased concentrations of Fe-binding organic ligands in the path of the TPD relative to adjacent sampling stations. These ligands and the associated Fe on the one hand were found to correlate strongly with terrestrial sources, which are projected to increase in a changing Arctic. On the other hand, Rijkenberg et al. (2018) found a local occurrence of Fe limitation over the Nansen basin and hypothesized that retreating ice could further exacerbate this nutrient limitation.

This paper is a synthesis of the distributions of TEIs in the central Arctic Ocean associated with the TPD. We examine the origin and fate of TEIs in this important trans-Arctic conduit and provide a first estimate of the mass transport rate for the TPD, based on ice drift trajectories and radionuclide tracers. By combining the TPD mass transport estimate with the TEI inventories reported herein, fluxes of these elements to the central Arctic Ocean via the TPD are estimated. Finally, we discuss the biogeochemical implications of the changing climate on TEI concentrations and fluxes to the Arctic and North Atlantic Oceans.

2 Study Area

The characteristics of water masses in the Arctic Ocean are controlled by bathymetry and inflows from the Atlantic and Pacific Oceans. The Arctic has two major basins, the Eurasian and Amerasian Basins, which are separated by the Lomonosov Ridge (Rudels, 2015). The Lomonosov Ridge is an underwater ridge of continental crust that emerges north of the Siberian shelves at approximately 140°E. Here we refer to the Amerasian Basin as the “western Arctic,” while the Siberian shelves and Eurasian Basin are referred to as the “eastern Arctic.” The Eurasian Basin is further divided into the Nansen and Amundsen Basins by the Gakkel Ridge, and the Amerasian Basin is divided by the Alpha-Mendeleev Ridge into the large Canada Basin and the Makarov Basin. Surrounding these basins are wide, shallow continental shelves that occupy over 50% of the Arctic Ocean's area (Jakobsson, 2002). Pacific water flows into the Arctic through the narrow and shallow Bering Strait, while Atlantic water enters through the Barents Sea and the Fram Strait (Rudels, 2009). The major outflows of Arctic waters are through the Canadian Arctic Archipelago and Fram Strait, on either side of Greenland, into the North Atlantic (Carmack et al., 2016) (Figure 1).

Figure 1

Map of the Arctic Ocean with the major upper ocean circulation features (blue arrows) as well as the Transpolar Drift (TPD; red arrows). The symbols indicate the station locations for the two GEOTRACES cruises GN01 (circles) and GN04 (triangles). The symbol colors denote the meteoric water fraction at each station. Also shown is the approximate location of the TPD origin for the positive (AO+) and negative (AO−) modes of the Arctic Oscillation (AO). The 200 m isobath is shown in gray.

Arctic Ocean sea surface pressure gradients are largely driven by water mass salinity differences in the basin. Between the (relatively fresh) North Pacific and the (salty) North Atlantic waters, there is a steric height gradient of about a meter, creating a pressure gradient across the Arctic from the Pacific down to the Atlantic. Large inputs of freshwater along the Arctic coastlines create a sea surface height gradient from the coasts to the central basins, which drives a series of boundary currents in the coastal seas and over the continental slope that move water eastward (counterclockwise) around the Arctic (Rudels, 2015; Rudels et al., 1994).

Overprinted on these perennial pressure gradients, the surface circulation is strongly impacted by winds. Predominant atmospheric circulation causes the average sea level pressure to be high over the Canada Basin and low over the Eurasian Basin, Barents Sea, and Nordic Seas (Hunkins & Whitehead, 1992; Serreze & Barrett, 2011). The resulting winds draw relatively fresh water over the Amerasian Basin and set up the anticyclonic Beaufort Gyre and a weaker cyclonic gyre in the Eurasian Basin (Alkire et al., 2015; Bauch et al., 2011; Carmack et al., 2016; Ekwurzel et al., 2001; Newton et al., 1974; Proshutinsky & Johnson, 1997). These two circulation cells converge just north of Siberia to form the Transpolar Drift (Rudels, 2015). The TPD extends from the Siberian shelves to the Fram Strait, as inferred from ice motion (Rigor et al., 2002) and water mass characteristics (McLaughlin et al., 1996).

The position of the TPD is determined by the Arctic Oscillation (AO), a large-scale Arctic climate pattern characterized by sea level pressure anomalies (Figure 1). The AO is highly correlated with the North Atlantic Oscillation (NAO) (Mysak, 2001), sea level pressure over the central Arctic, and with sea surface height anomalies along the coastal Arctic (Newton et al., 2006). During a low or negative AO and NAO, a strong Arctic High exists over the Canada Basin, expanding the anticyclonic Beaufort Gyre. In this case, the TPD originates from the Laptev and East Siberian Seas and flows over the Lomonosov Ridge (Morison et al., 2006; Woodgate et al., 2005) (solid red arrows in Figure 1). Positive AO and NAO indices produce a weak Arctic High, resulting in a smaller Beaufort Gyre (Mysak, 2001). In a persistently positive phase of the AO, the TPD shifts eastward toward the Bering Strait, entraining more Pacific water from the Chukchi Sea while still receiving a contribution from the East Siberian Shelf waters, which are transported farther east along the shelf before entering the TPD (Morison et al., 2012; Mysak, 2001) (dashed red arrows in Figure 1). During the years preceding the 2015 Arctic GEOTRACES sampling, the annual average AO was neutral to negative, and thus during the expeditions the TPD was located over the Lomonosov Ridge (Kipp et al., 2018; Rutgers van der Loeff et al., 2018). Monitoring of atmospheric circulation (Morison et al., 2012; Proshutinsky et al., 2009) as well as biogeochemical and water mass properties on previous hydrographic transects (Falck et al., 2005; Morison et al., 2012; Steele et al., 2004) provide evidence that this position has remained relatively stable over the past ca. 30 years.

The TPD is associated with high concentrations of dissolved organic matter (DOM) and a distinct oxygen isotope fingerprint relative to the surrounding surface waters. The DOM enrichment is due to river runoff and marine productivity occurring over the shelf where the TPD originates (Kaiser, Benner, et al., 2017; Stedmon et al., 2011). For the stable oxygen isotopes of water (δ¹⁸O-H₂O), negative oxygen isotope ratio anomalies are present in the core of the TPD—perhaps the lowest of all surface waters over the deep Arctic basins, between −3‰ and −4‰ (Bauch et al., 2011). The δ¹⁸O signal is a proxy for meteoric water (includes precipitation and runoff). In Arctic river deltas and estuaries the δ¹⁸O values are between about −18‰ and −22‰, while source waters in the Atlantic and Pacific have δ¹⁸O of approximately +0.3‰ and −1.1‰, respectively (Bauch et al., 2011; Ekwurzel et al., 2001; Newton et al., 2013). Salinity gradients across the TPD are smoother than δ¹⁸O changes, largely because the decrease in continental runoff outside the TPD is compensated by a rise in the amount of sea ice meltwater at the surface (Newton et al., 2013; Schlosser et al., 2002). These signals in DOM and δ¹⁸O are carried with the TPD all the way to the Fram Strait (Granskog et al., 2012).

The characteristics of the upper water column differ on either side of the TPD because it generally acts as a boundary between Atlantic and Pacific contributions to the Arctic pycnocline. High nutrient, high DOM, low-salinity Pacific water is typically observed as an “upper halocline” over the Canada and Makarov Basins, where it separates surface waters from the Atlantic boundary currents below about 200 m. Subsurface distributions of nitrate, phosphate and silicate indicate that a layer of nutrient-rich shelf-modified Bering Strait Inflow thins and shoals northward from the Chukchi continental slope and dissipates in the vicinity of the TPD. Pacific influence is dominant in the Canadian Arctic Archipelago (Jones et al., 2003; Jones & Anderson, 2008) and extends north of Greenland to the Fram Strait (de Steur et al., 2013; Dmitrenko et al., 2019). Over the Eurasian Basin, the Pacific-influenced layer is absent, with Atlantic waters occupying the entire water column (Bauch et al., 2011).

3 Methods

3.1 Sampling and Analyses of TEIs

The data presented herein were collected primarily during two cruises in 2015 associated with the Arctic GEOTRACES program. The U.S. GEOTRACES GN01 (HLY1502) cruise was held aboard the USCGC Healy, while the German GEOTRACES GN04 (PS94) cruise was on the R/V Polarstern (Figure 1). All sampling and analyses were conducted according to preestablished GEOTRACES approved protocols (for TEIs) (Cutter et al., 2014) and/or GO-SHIP approved protocols (for non-TEIs) (Hood et al., 2010). To further ensure quality of TEI data across participating laboratories, extensive intercalibration efforts were taken in accordance with GEOTRACES protocols (Cutter, 2013). For example, the GN01 and GN04 cruises both occupied the same station within 2 weeks of each other (GN01 station 30 and GN04 station 101), which enabled investigators to intercompare results for their respective TEIs. Conductivity Temperature Depthplease (CTD)/rosette data and methodologies for PS94 are available via the PANGEA database (Ober et al., 2016a, 2016b; Rabe et al., 2016a, 2016b). The GN01 CTD/rosette procedures are stored on the Biological and ChemicalOceanography Data Management Office (BCO-DMO) database (Landing et al., 2019a, 2019b). Detailed methodologies can be found in the publications where the original TEI data were first reported (Methods Appendix Table A1). Certain TEIs were analyzed by multiple laboratories using similar or independent methods. In these cases, we used the average value for each station and depth. The methods and data appendices provide more specifics on data averaging and sources.

3.2 Linear Mixing Model

In order to study the provenance and pathways of TEIs carried by the TPD, we must quantitatively parse the fraction (f) of source waters in each collected GEOTRACES sample. To do so, we use the relatively well-studied distribution of salinity (S), δ¹⁸O-H₂O ratios, and the Arctic N-P tracer (ANP; see Newton et al., 2013). These can be used to identify fractions of Pacific (Pac)- and Atlantic (Atl)- sourced seawater, sea ice melt (SIM), and meteoric water (Met). The latter includes runoff and net in situ precipitation. Along the cruise transects, in situ precipitation is expected to be small in comparison with the continental runoff; hence f_Met will be our primary proxy for determining the water masses most influenced by the TPD. The value for each in a sample is expressed as a linear combination of the values in its constituent water masses:

This constitutes a four-dimensional linear system that can be solved in matrix form:

where [f] is a vector of water mass fractions, [y] is a vector of the parameter values in the sample, and {C} is a matrix of values in the “end-members,” that is, the source waters. The model assumes four end-members (Table 1) and four equations, so will yield an exact solution.

Table 1. End-Member Parameter Values for the Water Mass Analysis Linear Mixing Model

Water mass	Salinity	δ^18O (‰)	Arctic N:P^a^,^b
Atlantic water	34.92	+0.3	0
Pacific water	32.50	−1.1	1
Meteoric water	0	−19	0
Sea ice meltwater	4	Surface + 2.6‰	Surface

• a Newton et al. (2013).
• b Pacific water: slope = 14, intercept = −11; Atlantic water: slope = 17, intercept = −2.

There are several important sources of error, which are discussed in detail by Newton et al. (2013) in the context of the 2005 Arctic Ocean Section. Briefly, the least constrained fractions are those of Pacific- and Atlantic- influenced ocean water, which suffer from the nonconservative nature of nutrients in the ocean, large scatter in the values in the source waters, and potentially from drift in the end-member means with time (Newton et al., 2013). Fortunately, our focus here is on the concentration of meteoric waters and this fraction is insensitive to nutrient concentrations. Rather, it depends on salinity and δ¹⁸O with the error originating primarily from seasonal and geographical variability in the δ¹⁸O end-member of Arctic rivers (Cooper et al., 2008). Monte Carlo analysis across a reasonable range of estimated mean δ¹⁸O values for runoff yielded (one-sigma) errors of about 1% (absolute value) on the meteoric fractions.

The relationship between each TEI and the meteoric water fraction was determined using a linear regression model. The slope, intercept, r² value, and p value for each relationship are reported in Table 2. The effective shelf end-member concentrations of the TEIs were calculated using their respective linear regressions at 20% meteoric water, assuming that this is the meteoric water fraction of the TPD when it leaves the shelf and that there was no significant TEI removal or addition during transport. Meteoric water fractions of 10–35% have been observed at the point of origin of the TPD in the Laptev Sea (Bauch et al., 2011) and its terminus at the Fram Strait (Dodd et al., 2012). During the 2015 GEOTRACES expeditions, fractions up to 25% were observed near the North Pole, thus 20% is a conservative estimate.

Table 2. Linear Curve Fit Data and Statistics for the TEI Versus Meteoric Water Relationship Plots

Property	Slope	y-intercept	r^2	p
Alk (μM)	−14.1788	2,365.40	0.53	0.000
CDOM (V)	0.0038	0.08	0.59	0.000
cFe (dFe-sFe) (nM)	0.0924	−0.83	0.88	0.000
d232Th (pM)	0.0941	−0.56	0.83	0.000
dAl (nM)	0.0501	0.55	0.23	0.093
dBa (nM)	1.1597	41.61	0.68	0.000
dCd (nM)	0.0117	0.05	0.66	0.000
dCo (pM)	7.9853	55.78	0.54	0.000
dCo_labile (pM)	2.6214	−19.11	0.61	0.008
dCu (nM)	0.2881	1.25	0.96	0.000
del_114Cd (‰)	0.0077	0.38	0.10	0.319
del_56Fe (‰)	0.0504	−0.82	0.51	0.009
del_66Zn (‰)	0.0220	0.08	0.14	0.227
dEr (pM)	0.2492	7.34	0.49	0.000
dFe (nM)	0.1960	−0.33	0.67	0.000
dGa (pM)	−0.3859	14.93	0.58	0.000
DIC (μM)	−11.4712	2,240.12	0.43	0.000
dMn (nM)	0.1509	1.31	0.41	0.000
dNd (pM)	0.9532	24.73	0.54	0.000
dNi (nM)	0.2750	3.09	0.91	0.000
DOC (μM)	3.9774	52.91	0.88	0.000
dPb (nM)	−0.0005	0.01	0.59	0.000
dV (nM)	−0.6114	25.70	0.65	0.001
dZn (nM)	0.0525	0.28	0.36	0.001
Total Hg (pM)	0.0061	1.15	0.01	0.600
Total MeHg (pM)	−0.0055	0.16	0.28	0.007
Nitrate (μM)	−0.2432	4.97	0.48	0.000
pAl (nM)	−0.6779	23.08	0.01	0.521
pFe (nM)	−0.1068	4.96	0.01	0.613
Phosphate (μM)	0.0124	0.46	0.13	0.000
pMn (nM)	−0.0062	1.84	0.00	0.895
Ra226 (dpm/100 L)	0.2523	5.79	0.49	0.000
Ra228 (dpm/100 L)	0.9614	1.10	0.81	0.000
sFe (<0.02μm) (nM)	0.1778	−1.22	0.79	0.000
Silicate (μM)	0.4576	3.35	0.30	0.000
POC	0.0188	0.20	0.18	0.342
d13C POC (‰)	0.0912	−31.10	0.04	0.665
p bSi (μM)	0.0027	−0.02	0.40	0.128

Initial estimates of river end-member concentrations were calculated by extrapolating the linear regression to 100% meteoric water (regression intercept). These estimates have a high statistical uncertainty associated with them due to the extrapolation beyond the measured range and other factors that violate the assumptions of the standard estuarine mixing model (Boyle et al., 1974; Shiller, 1996), but they still provide a first approximation to compare with sparse existing river and shelf sea data. There is some data on TEI concentrations in the Eurasian rivers that ultimately feed into the TPD. Most are derived from the Arctic Great Rivers Observatory (A-GRO), which began as the Pan-Arctic River Transport of Nutrients, Organic Matter and Suspended Sediments (PARTNERS) project (Holmes et al., 2019). The weighted averages reported by the A-GRO provide a useful comparison for many of the elements discussed in this manuscript, but could be improved with measurements of more TEIs in each of the Arctic rivers and knowledge of the relative influence of each river in the TPD at a given time. Due to the shelf circulation patterns (Figure 1), the major Eurasian rivers (Lena, Ob', Yenisey, and Kolyma) will exert a stronger influence on the TPD than the North American rivers (Mackenzie and Yukon). As such, we report herein the discharge weighted average TEI concentrations for the Eurasian rivers only. Most importantly, any differences between the effective river end-member and the mean river concentrations should not be interpreted in a quantitative manner; rather, this analysis is meant only to give the reader a sense of the relative influence of rivers and/or estuarine removal/addition processes on the TEIs that are transported to the central Arctic Ocean via the TPD.

4 Results and Discussion

We define the lateral extent of the TPD as ~84°N (in the Canada Basin) to 87°N (in the Eurasian Basin) for waters in the top 50 m. These boundaries were chosen qualitatively based on the distributions of the meteoric water fraction and TPD-influenced TEIs (Figure 2). For example, there is a sharp concentration gradient for chromophoric dissolved organic matter (CDOM), dissolved organic carbon (DOC), dissolved Fe, and ²²⁸Ra at stations north of 84°N 150°W, which coincides with a front between high and intermediate meteoric water fractions (~250 km along the section distance in Figure 2). On the Eurasian side of the transect, there is minimal meteoric water influence south of 87°N (~1,100 km along the section distance in Figure 2). The TPD can be characterized generally by this high meteoric water component, which is due to large river contributions to the Siberian Arctic shelves. However, the meteoric water fraction alone cannot be used to delineate the western boundary of the TPD because the Beaufort Gyre in the Canada Basin contains a significant and growing freshwater component sourced from eastern Arctic rivers (Giles et al., 2012; Morison et al., 2012; Rabe et al., 2011, 2014).

Figure 2

Section plots for (a) % meteoric water, (b) salinity, (c) DOC, (d) dFe, (e) CDOM, (f) ²²⁸Ra, (g) silicate, and (h) dNd along a transect that spans two GEOTRACES cruises and bisects the Transpolar Drift. The stations included in the contour plots is shown on the map inset for (g), and the distance is relative to a station located at 82°N, 150°W.

As a function of depth, the elevated TEI concentrations and meteoric water fractions are confined to the upper 50 m. The 50 m cutoff also serves to exclude the halocline from our analysis, which is rich in certain TEIs and nutrients like silicate (Figure 2g), and is influenced by different ventilation processes and source water masses than the TPD (Aagaard et al., 1981). The data presented herein are shown mainly as a function of the meteoric water fraction and were collected in the upper 50 m of the water column for all stations north of 84°N, which includes the polar mixed layer and the TPD.

4.1 Trace Element and Isotope Distributions, Sources, and Sinks

4.1.1 Nutrients

Within the Transpolar Drift influenced stations, nitrate concentrations generally decrease with increasing meteoric water fraction, while phosphate and silicate displayed the opposite trend (Figures 3a–3c). This pattern is in contrast with the average Eurasian river nitrate of 4.2 μmol/L (Holmes et al., 2019) and suggests that nitrate is largely assimilated (Arrigo et al., 2008) or consumed by denitrification (Chang & Devol, 2009) over the shelf prior to entering the TPD, and acts as the limiting nutrient for primary production, leaving residual phosphate and silicate concentrations of ~0.75 and 12 μmol L⁻¹, respectively.

Figure 3

Trace element and isotope concentrations (a) nitrate, (b) phosphate, (c) silicate, (d) DOC, (e) CDOM, (f) DIC, (g) alkalinity, (h) dBa, (i) ²²⁸Ra, (j) ²²⁶Ra, and k) ²³²Th as a function of the meteoric water percentage at stations from the GEOTRACES GN01 (red circles) and GN04 (blue circles) cruises. Data are restricted to 0–50 m for stations north of 84°N. Regression lines are shown for variables with significant (p < 0.05) relationships with meteoric water.

A linear fit to the silicate data and extrapolation to 100% meteoric water suggests an apparent riverine end-member concentration of 47 μmol L⁻¹. This estimate can be compared to the average discharge weighted silicate concentration for the Eurasian rivers reported by the A-GRO, 83 μmol L⁻¹ (Holmes et al., 2019). The apparent silicate riverine end-member is slightly more than half of the A-GRO weighted average, suggesting removal in estuaries or over the shelf, although shelf sediments can also act as a source of additional silicate (Frings, 2017; Tréguer & De La Rocha, 2013).

Given the large phosphorous (P) fluxes through the Arctic gateways, rivers are thought to be of relatively minor importance in the Arctic Ocean P budget (Holmes et al., 2012). Their P relationship with meteoric water fraction was relatively weak, which precludes us from estimating the effective river end-member P concentration. For comparative purposes only, we note that the P concentration range at the highest observed meteoric water percentage (>20%) is ~0.6–0.7 μmol L⁻¹, which is about two times higher than the weighted Arctic Eurasian river phosphate average of 0.31 μmol L⁻¹ (Holmes et al., 2019). This finding is consistent with the study by Torres-Valdés et al. (2013), which suggested that the Arctic Ocean P budget imbalance (excess) cannot be explained entirely by riverine inputs.

The Arctic Ocean is at present known to be a net exporter of Si and P to the North Atlantic Ocean, where the excess P is thought to be partly responsible for N-fixation (Torres-Valdés et al., 2013). Looking toward the future, it is unclear if additional terrestrial inputs of nutrients will lead to an additional flux of nutrients to the TPD (and beyond) because some fraction would be consumed over the productive shelves before being exported toward the central basin. Further, though the mass flux of nutrients from rivers is small relative to the influx of nutrients to the Arctic through the Pacific and Atlantic gateways (Holmes et al., 2012), the river-influenced TPD is a nutrient source to the surface ocean where it is immediately available for use by phytoplankton. Whether or not the TPD becomes a more significant source of nutrients to Arctic primary production will depend on changes in light availability (due to ice loss; Ji et al., 2013; Nicolaus et al., 2012), stratification, and extent of denitrification in Arctic shelf sediments.

4.1.2 Dissolved Inorganic Carbon and Alkalinity

The surface water (<50 m) concentrations of dissolved inorganic carbon (DIC;1920–2,260 μmol L⁻¹) and total alkalinity (TA; TA;1,990–2,340 μmol L⁻¹; Figures 3f and 3g) largely reflect the salinity distribution and gradient (27.2–34.4) north of 84°N. River water contributes to Arctic Ocean DIC and TA, while sea ice meltwater will result in a dilution of these two species. Waters with the largest meteoric fraction (20–23%) are characterized by slightly higher concentrations of DIC and TA compared to corresponding salinity values of 28 in the southern Eurasian Basin that are beyond the influence of rivers.

There are negative linear correlations between the meteoric water fraction and concentrations of DIC (r² = 0.43, p < 0.001) and TA (r² = 0.53, p < 0.001) in the upper 50 m in samples north of 84°N. The lower DIC (1,920–2,230 μmol L⁻¹) and TA (2,020–2,100 μmol L⁻¹) and large fractions of meteoric water (>20%) are associated with high concentrations of DOC (120–150 μmol L⁻¹), reflecting the input of Siberian shelf water and river runoff to the central basins. The effective river end-members for DIC and TA based on extrapolation of the linear fit to 100% meteoric water are 1,090 and 950 μmol L⁻¹, respectively. These fit well with the discharge weighted concentrations of these two parameters for the major Arctic rivers as reported by Tank et al. (2012; DIC = 1,110 μmol kg⁻¹; TA = 1,010 μmol kg⁻¹) or Eurasian rivers only (TA = 800 μmol kg⁻¹, Cooper et al., 2008; TA = 815 μmol kg⁻¹, Holmes et al., 2019).

Seasonally varying discharge from the major Siberian rivers drives much of the variation in the surface distribution of DIC and TA in the shallow shelf seas (Drake et al., 2018; Griffin et al., 2018). As the Arctic warms and the permafrost thaws, the hydrologic cycle is accelerating and the total river discharge is increasing (Griffin et al., 2018), resulting in increased river export of nutrients, DOC, DIC, and TA (Drake et al., 2018; Kaiser et al., 2017; Pokrovsky et al., 2015; Tank et al., 2016). The Siberian shelf seas experience increasingly ice-free conditions (e.g., Serreze et al., 2007) and are areas of extensive biogeochemical transformation of organic matter, of both marine and terrestrial origin. This extended ice-free condition, in combination with brine production from sea ice formation, results in a cold bottom water of relatively high salinity and partial pressure of carbon dioxide (pCO₂). This high-pCO₂ water is partly outgassed to the atmosphere, and partly distributed on the outer shelf, as well as into the central Arctic basins depending on season, sea ice conditions, and wind field (Anderson et al., 2009; Anderson, Björk, et al., 2017).

Degradation of terrestrial organic matter and substantial discharge of Arctic river water with elevated pCO₂ leads to persistently low pH in the Siberian shelf seas. This calcium carbonate corrosive water has been observed all along the continental margin and well out into the Makarov and Canada basins, though the effects are most pronounced below the terrestrially influenced upper layer, at depths between 50 and 150 m (Anderson, Ek, et al., 2017; Cross et al., 2018). This feature coincides with high nutrient concentrations of the upper halocline waters, consistent with organic matter remineralization as described by, for example, Jones and Anderson (1986). This layer also holds the highest levels of pCO₂ (up to 780 μatm) and, consequently, the lowest values of the saturation state of the calcium carbonate polymorph aragonite Ω_AR (down to 0.7). Hence, the TPD is an increasingly important feature in the distribution of physical and biogeochemical properties in general, and in the seawater CO₂ system and ocean acidification in particular.

In the TPD north of 84°N, Ω_AR calculated from TA and pH_SWS^25C at in situ temperature and pressure, is effectively at saturation (Ω_AR = 1.06 ± 0.07, n = 17). Increasing river discharge, shelf remineralization of organic matter, and shelf-basin interactions are expected to further decrease the saturation state in the surface layers of the central Arctic in the future (Anderson, Björk, et al., 2017; Brown et al., 2016; Semiletov et al., 2016; Wynn et al., 2016).

4.1.3 Dissolved Organic Matter

The average DOC concentration of Atlantic water entering the Arctic Ocean is 60 μmol L⁻¹ at the surface and less than 50 μmol L⁻¹ in the intermediate and deep waters (Amon et al., 2003; Anderson & Amon, 2014). This Atlantic water dominates the water column in the central Arctic Ocean, with its core in the upper 500 m. In the western Arctic Ocean, Pacific water (70 μmol L⁻¹) prevails in the top 100 m. Both Pacific and Atlantic waters are characterized by low CDOM (Anderson & Amon, 2014). Primary production in the Arctic Ocean contributes to the DOM distribution in surface waters where a seasonal signal can be observed, especially in open water over the productive shelves (Davis & Benner, 2005; Mathis et al., 2007).

Of all the characterized sources of DOC to the Arctic Ocean, river runoff has the highest concentration of DOC (350–990 μmol L⁻¹; Amon et al., 2012) along with elevated levels of CDOM (Anderson & Amon, 2014; Stedmon et al., 2011). In the eastern Arctic Ocean, DOC and CDOM have larger components of terrigenous DOM (Amon, 2004; Amon et al., 2003; Benner et al., 2005; Kaiser, Benner, et al., 2017), whereas DOC and CDOM in the Canada Basin have lower concentrations of terrigenous relative to marine-derived DOM (Benner et al., 2005; Stedmon et al., 2011). Fluvial discharge entrained in the TPD is a major source of terrigenous DOC and CDOM to surface waters of the central Arctic (Amon, 2004; Kaiser, Canedo-Oropeza, et al., 2017; Letscher et al., 2011; Opsahl et al., 1999; Shen et al., 2016).

Sea ice processes add to the complexity of DOC, CDOM and meteoric water distributions in the surface waters of the Arctic Ocean. Ice melt into low DOM ocean waters can stimulate primary production of DOM or be a source of ice algae DOM, but would not add CDOM to surface waters (Anderson & Amon, 2014). In contrast, sea ice melt results in a dilution of surface water DOM (Granskog et al., 2015), but can add meteoric water from snow and river water included in sea ice. Within the TPD, the relationship between the meteoric water fraction and the DOC concentration was stronger (r² = 0.88, p < 0.001) than the relationship between the meteoric water and CDOM (r² = 0.59, p < 0.001) (Figures 2c, 2e, 3d, and 3e) supporting the notion that DOC and CDOM are independently controlled by different processes. Biological processes influence DOC and CDOM to different degrees, primary production will increase DOC but not CDOM, microbial degradation will decrease DOC, but might actually increase CDOM. Physicochemical processes like photobleaching and flocculation will affect CDOM but not necessarily DOC while freezing affects DOC and CDOM in a similar fashion (Guéguen et al., 2012; Kaiser, Canedo-Oropeza, et al., 2017; Moran et al., 2000; Sholkovitz, 1976; Uher et al., 2001).

The effective meteoric water end-member for the linear regression with DOC is 451 μmol L⁻¹, which is about half of the discharge weighted value for the major Eurasian Arctic rivers (800 μmol L⁻¹; Holmes et al., 2012, 2019). Similar losses were reported for CDOM (Granskog et al., 2012), and both reflect removal during estuarine mixing and passage over the shelf seas by flocculation, photomineralization or microbial degradation (see section 4.1.2) (Alling et al., 2010; Hansell et al., 2004; Kaiser, Canedo-Oropeza, et al., 2017; Letscher et al., 2011), or the influence of shelf ice melt as described by Amon et al. (2012). Earlier studies have reported conservative mixing of DOC in Arctic river estuaries (Amon & Meon, 2004; Köhler et al., 2003), but they were based on late summer sampling and excluded the freshet period, which delivers more biolabile DOM to the Arctic coast.

Regarding the marine production of DOC and CDOM, warming temperatures and a decrease in ice cover will result in increased primary productivity, since light is the major limiting factor for the production of organic carbon by phytoplankton in the Arctic Ocean (Vancoppenolle et al., 2013). An Arctic Ocean-wide annual 20% increase in net primary production has already been reported based on remote sensing ocean color products (Arrigo & van Dijken, 2011) and will translate into more DOM derived from decomposing planktonic sources. As CDOM absorbs ultraviolet and visible light, the expected higher fluvial CDOM fluxes along with increased marine CDOM might have a negative feedback effect on the light limitation in the shelf areas (Pavlov et al., 2015) but should facilitate TEI transport to the open Arctic Ocean for those particle-reactive trace elements that form stable complexes with DOM (e.g., Fe, Cu, Ni, and Co).

4.1.4 Particulate Matter

In the upper 50 m of TPD-influenced stations, none of the particulate (p) trace elements and isotopes (pTEIs) examined here show linear correlations with the fraction meteoric water (Figure 4). While the lithogenic elements such as pFe and pAl were sometimes enriched by over an order of magnitude relative to their dissolved concentrations in the TPD (Figures 4d–4f), they were not statistically correlated with the meteoric water fraction (Figures 4a–4c).

Figure 4

Particulate trace element and isotope concentrations or particulate/dissolved ratios (a) pFe, (b) pAl, (c) pMn, (d) pFe/dFe, (e) pAl/dAl, (f) pMn/dMn, (g) POC, (h) p bSi, and (i) as a function of the meteoric water percentage at stations from the GEOTRACES GN01 (red circles) and GN04 (blue circles) cruises. Data are restricted to 0–50 m for stations north of 84°N. The GN01 data are from the small size fraction (1–51 μm) in situ pump filters, while the GN04 samples were bottle collected on 0.45 μm filters. Regression lines are shown for variables with significant (p < 0.05) relationships with meteoric water.

The particulate organic carbon (POC) and biogenic silica (bSi; Figures 4g and 4h) also did not show significant correlations with meteoric water. However, only surface data from GN01 are available for these variables. Since GN01 stations did not span a large dynamic range in meteoric water, it is unclear whether the TPD could be a source of riverine POC to the central basin in the same way as it is for DOC. The δ¹³C-POC within the TPD is extremely depleted (Figure 4i), potentially consistent with an influence of depleted (−30‰) riverine organic matter that is of terrestrial origin (Holmes et al., 2019; McClelland et al., 2016), but it could also be explained by the large isotope fractionation observed in slow growing phytoplankton (Brown et al., 2014). The residence time of small size fraction (SSF; 1–51 μm) POC in the upper 100 m within the TPD is about 300 days at 88°N (station 38), and even longer at the other TPD stations (Black, 2018). This extremely low rate of particle loss in the upper Arctic Ocean (Black, 2018) is of similar magnitude to the time scale of TPD transport from the shelf to the central Arctic (Kipp et al., 2018), potentially allowing for riverine POC to survive transport to the central Arctic.

In the context of rapidly changing climate, as permafrost thaws, river flux and coastal erosion increases, higher concentrations of riverine POC could be transported into the central Arctic Basin. There, the POC may be subject to more intense microbial degradation due to the expected rise in ocean temperature (Kirchman et al., 2005; Middelboe & Lundsgaard, 2003). Furthermore, the TPD is transporting younger sea ice that is subject to melting before it reaches Fram Strait, thereby increasing central Arctic accumulation of ice-rafted material, such as POC, and decreasing export to the Atlantic (Krumpen et al., 2019).

4.1.5 Radium Isotopes and Barium

Radium-228 (t_1/2 = 5.75 years) activities were high in the TPD and had a strong positive correlation with the fraction of meteoric water (r² = 0.81, p < 0.001). Activities increased from <5 dpm 100 L⁻¹ at meteoric water fractions <5% to 20–25 dpm 100 L⁻¹ at meteoric water fractions ~20% (Figure 3i). These high ²²⁸Ra activities persisted over the upper 50 m of the water column. The correlation between ²²⁶Ra (t_1/2 = 1,600 years) and meteoric water (r² = 0.49, p < 0.001) was not as strong as that for ²²⁸Ra, but ²²⁶Ra levels did increase from ~6–8 dpm 100 L⁻¹ at low meteoric water fractions to ~9–11 dpm 100 L⁻¹ in the core of the TPD (Figure 3j). Radium-226 activities of samples collected on the western side of the Lomonosov Ridge remained high even outside of the region influenced by the TPD, due to the influence of high-²²⁶Ra Pacific inflow in the Canada Basin. In contrast, activities decreased (<6 dpm 100 L⁻¹) in the Atlantic-influenced Eurasian Basin.

The strong correlation between ²²⁸Ra and the fraction of meteoric water reflects the shelf signal carried in the TPD. While ²²⁸Ra does have a riverine source (Rutgers van der Loeff et al., 2003), shelf sediments supply over 80% of the ²²⁸Ra in Arctic surface waters (Kipp et al., 2018). The weighted average annual ²²⁸Ra activity of Arctic rivers is on the order of 24 ± 13 dpm 100 L⁻¹; desorption of Ra from suspended particles in the estuarine mixing zone could add an additional 25% (Kipp et al., 2018). This combined riverine ²²⁸Ra source is significantly lower than the effective river end-member (97 dpm 100 L⁻¹) determined by its relationship with meteoric water. This correlation therefore results from the transport of river water over the shallow Eurasian shelves before the TPD carries this river- and shelf- influenced signal to the central Arctic. The activities of ²²⁸Ra measured in the TPD were higher in 2015 compared to 2007 and 2011, indicating an increased flux of ²²⁸Ra to the central Arctic (Kipp et al., 2018; Rutgers van der Loeff et al., 2018). This rise is likely driven by the loss of ice cover over Eurasian shelves, permitting increased wind-driven vertical mixing that can transport ²²⁸Ra produced in shelf sediments into the overlying water column through enhanced sediment resuspension and porewater exchange (Kipp et al., 2018; Serreze et al., 2007; Williams & Carmack, 2015).

The weaker correlation between ²²⁶Ra and meteoric water results from the larger surface water inventory and smaller shelf source of this isotope compared to ²²⁸Ra. Once removed from shelf sediments (through diffusion or resuspension) ²²⁶Ra will regenerate more slowly from decay of its Th parent than ²²⁸Ra due to its longer half-life. The ratio of ²²⁸Ra/²²⁶Ra inputs from shelves is therefore typically greater than 1, and ratios as high as 3.9 have been observed on the Laptev Shelf (Rutgers van der Loeff et al., 2003). On the GN01 and GN04 transects, ²²⁸Ra/²²⁶Ra activity ratios were between 1 and 2 at meteoric water fractions >15%, reflecting the high shelf ratios of these isotopes carried in the TPD. While activities of ²²⁶Ra in the TPD have increased from 2007 to 2015, this change was not as large as that for ²²⁸Ra (Kipp et al., 2018; Rutgers van der Loeff et al., 2018). The longer half-life of ²²⁶Ra compared to ²²⁸Ra results in a larger surface water inventory, thus a substantial rise in inputs is required to increase surface water ²²⁶Ra activities.

Barium is a chemical analog of radium. The highest surface dBa concentrations (<25 m) are observed in the Canada Basin (67.7 ± 1.4 nmol L⁻¹; range: 64.4–69.1 nmol L⁻¹). Makarov Basin surface waters are slightly lower (62.5 ± 0.7 nmol L⁻¹; range: 53.4–63.3 nmol L⁻¹), but are still high relative to incoming Atlantic and Pacific seawater. These high concentrations likely indicate the presence of accumulated river water circulating in the Amerasian Basin (Guay et al., 2009). Samples in the Nansen Basin and the Barents Sea are lower and representative of the Atlantic source (~40 nmol L⁻¹). However, data from the Amundsen Basin are more variable, with surface concentrations that range from 48.6–65.5 nmol L⁻¹. This variation appears to be driven by the composition of the water at each station: where there are high ice melt or Atlantic fractions, the concentrations are lower (<60 nmol L⁻¹). Alternatively, where Pacific water or meteoric fractions are relatively high, the concentrations are higher (>60 nmol L⁻¹).

There is a strong positive correlation between dBa concentration and % meteoric water in the TPD (r² = 0.68, p < 0.001; Figure 3h), which is driven by high river Ba concentrations (Abrahamsen et al., 2009; Guay & Falkner, 1997, 1998). Scatter around the trend may result from nonconservative behavior of Ba such as removal from the dissolved phase in association with biological activity (Pyle et al., 2018, 2019), particularly over the productive Arctic shelf regions (Roeske et al., 2012). Further complicating matters, the shelf may also be a source of Ba, as was demonstrated for ²²⁸Ra (Kipp et al., 2018).

The best fit trendline for the dBa versus meteoric water relationship suggests an effective river end-member concentration of 158 nmol L⁻¹, which is higher than the discharge weighted dBa concentrations from the Eurasian rivers (92 nmol L⁻¹) (Holmes et al., 2019). Importantly, the data available from the A-GRO network are direct measurements of dBa in surface freshwater (S < 1) and since Ba is known to undergo desorption from suspended particulate matter (SPM) in the estuarine zone as salinity increases (e.g., Coffey et al., 1997), these measurements may not be representative of the effective end-member for riverine dBa. Accounting for desorption of Ba from SPM, Guay et al. (2009) reported effective end-members between 100 and 130 nmol L⁻¹ dBa in Eurasian rivers, which is closer to the value we determined from the linear regression with meteoric water fraction in the TPD. Supply of Ba from multiple rivers to the eastern Arctic shelves may also explain the slightly lower r² for the fit to meteoric water fraction as compared to other TEIs reported herein. Additionally, submarine groundwater discharge (SGD) can be a source of Ba to the marine environment as it generally contains high dBa (e.g., Shaw et al., 1998); at present, there are no constrained SGD dBa end-member concentrations available for the Arctic, but recent groundwater dBa measurements by Kipp et al. (2020) of 610–970 nmol L⁻¹ suggest that this source may need to be taken into account for Arctic Ba geochemical budgets moving forward.

As Arctic temperatures continue to rise, Ra and Ba inputs to Arctic surface waters will also be amplified by permafrost degradation, increased river discharge, and increased coastal erosion. As permafrost thaws, SGD is likely to become a more important part of the Arctic hydrologic cycle (Walvoord et al., 2012), and will serve as an additional source of these TEIs (Charkin et al., 2017). The combination of thermal erosion of permafrost and physical erosion from more turbulent shelf seas will also result in more coastal sediment delivery to Arctic shelf seas, which can release Ra isotopes and Ba through desorption.

Notably, dBa (and ²²⁶Ra to a lesser degree) can be influenced biologically and may be removed as barite precipitates, scavenged by particle surfaces, or taken up to some degree by phytoplankton cells (Bishop, 1988; Dehairs et al., 1980; Roeske et al., 2012). This nonconservative behavior complicates the element's utility as a tracer. As such, the influence of higher biological/particle interactions on Ba cycling in the Arctic as the central basins lose ice cover is difficult to anticipate.

4.1.6 Dissolved Trace Metals and Metal Isotopes

Along GN01, 11 dissolved (<0.2 μm) trace metals (Al, V, Mn, Fe, Co, Ni, Cu, Zn, Ga, Cd, and Pb) were evaluated, as well as the size partitioning of dissolved Fe (dFe) into soluble (sFe < 0.02 μm) and colloidal (0.02 μm < cFe < 0.20 μm) size fractions. The fraction of the total dCo that was chemically labile (LCo) versus strongly organically complexed was also quantified. Several metals displayed a concentration enrichment associated with the TPD in the central Arctic, as well as a significant relationship (95% confidence interval) with meteoric water north of 84°N and within the upper 50 m. Based on the r² values for a linear fit to the data, changes in meteoric water loadings explained >50% of the dFe, dCo, LCo, dNi, dCu, dCd, dGa, and dPb concentration variability and >40% of dMn variability (Figures 5a, 5f–h, 5l, and 6c–e). Dissolved Zn had a significant relationship (p = 0.001) with meteoric water in the TPD, but a comparatively lower r² (0.36; Figure 5j), while dissolved Al with meteoric water displayed no statistical significance at all (Figure 6a). Dissolved V displayed a strong negative correlation with meteoric water (Figure 6b; r² = 0.65, p = 0.001).

Figure 5

Trace element and isotope concentrations (a) dFe, (b) colloidal Fe, (c) soluble Fe, (d)as a function of the meteoric water percentage at stations from the GEOTRACES GN01 (red circles) and GN04 (blue circles) cruises. Data are restricted to 0–50 m for stations north of 84°N. Regression lines are shown for variables with significant (p < 0.05) relationships with meteoric water.

Figure 6

Trace element and isotope concentrations (a) dAl, (b) dV, (c) dGa, (d) dCo, (e) labile Co, (f) total Hg, (g) MeHg, (h) dEr, and (i) dNd as a function of the meteoric water percentage at stations from the GEOTRACES GN01 (red circles) and GN04 (blue circles) cruises. Data are restricted to 0–50 m for stations north of 84°N. Regression lines are shown for variables with significant (p < 0.05) relationships with meteoric water.

The measured dissolved metal concentrations are consistent with previous, albeit limited, literature on trace metals from the Arctic Ocean. Early studies of some trace metals such as Mn, Fe, Ni, Cu, Zn and Cd demonstrated low surface concentrations, with enrichments between 100 and 200 m in the halocline, and low but uniform values in the deep ocean (Danielsson & Westerlund, 1983; Moore, 1981; Yeats, 1988; Yeats & Westerlund, 1991). More recent studies focused primarily on the Chukchi Shelf (Aguilar-Islas et al., 2013; Cid et al., 2012; Kondo et al., 2016; Nishimura et al., 2012) or on the Eurasian Basin (Klunder, Bauch, et al., 2012; Klunder, Laan, et al., 2012; Middag et al., 2011), but again show very similar patterns, particularly with regard to the Canada Basin halocline feature that dominates many of the trace metal distributions. Within the TPD region, dFe and dMn concentrations compare well to those measured previously in studies focused on the Eurasian side of the Arctic Ocean (Klunder, Bauch, et al., 2012; Klunder, Laan, et al., 2012; Middag et al., 2011).

Compared to other global ocean surface waters in the 2017 GEOTRACES Intermediate Data Product (Schlitzer et al., 2018), Arctic Ocean surface waters have anomalously high concentrations of dissolved trace metals such as Fe, Co, Zn, Ni, Cu, and Cd, likely due to a combination of larger external sources and less biological uptake and/or scavenging removal under the sea ice (Figure 7). This surface enrichment can be seen for dFe (0.2 to 4 nmol L⁻¹) and dCu (3.9 to 7.4 nmol L⁻¹) in comparison to measurements from the Atlantic Ocean at a similar distance from the continental margin (Figure 7). While the wide continental shelves, sea ice, dust, and incoming Pacific and Atlantic waters are all potential sources of trace metals to surface Arctic waters, the strong correlation between meteoric water and dFe, dCo, dCu, and dNi north of 84°N point to a riverine source that originated in the eastern Arctic Ocean. However, since the meteoric water signal of the TPD is significantly modified and transformed in the Laptev Sea (Kadko et al., 2019; Kipp et al., 2018), concentrations of certain dissolved metals in TPD waters over the North Pole can only be interpreted when high-particle shelf and estuarine reactions are considered. For example, scavenging-prone metals had an inverse (dGa, dPb) or very poor relationship (dAl) with meteoric water over the central Arctic, despite high concentrations of these trace metals in Arctic rivers (e.g., Pb of up to 480 pmol L⁻¹ in the Lena River; Colombo et al., 2019; Hölemann et al., 2005). In Kadko et al. (2019), it is suggested that scavenging on the continental shelves may be a major sink of dissolved Pb from the Arctic Ocean. These lines of evidence suggest that removal of these TEIs by flocculation in estuaries and/or by scavenging onto the abundant particles over the Laptev Sea continental margin are important processes that greatly limit their being transported with the TPD into the central Arctic.

Figure 7

Surface to 600 m trace element and isotope concentration profiles for Arctic GEOTRACES GN01 station 32 (90°N; gray diamonds) as compared to Atlantic GEOTRACES GA03 station 8 (35.4°N, 66.5°W; black circles). (a) dFe, (b) dMn, (c) dAl, (d) dCu, and (e) ²³²Th.

Dissolved Fe and Mn are expected to undergo similar scavenging and flocculation processes as dAl, and so despite their extremely high concentrations in the Eurasian Arctic rivers (averages of 2,300 and 470 nmol L⁻¹, respectively; Holmes et al., 2019), one would expect lower concentrations downstream in the TPD. Indeed, concentrations of dFe and dMn in the TPD were between two and three orders of magnitude lower than river concentrations (2.90 ± 1.0 nmol L⁻¹ and 4.2 ± 1.4 nmol L⁻¹, respectively). Dissolved Mn had a lower degree of correlation with meteoric water and was present at concentrations similar to the central Atlantic (Figure 7b), suggesting that it was oxidized continually to the particulate phase and/or scavenged during transit across the Arctic. In contrast, dFe maintained a strong linear relationship with meteoric water despite the massive loss of Fe between rivers and the central Arctic. Notably for dissolved Fe, both smaller sFe and larger cFe had strong relationships with the fraction of meteoric water (Figures 5b and 5c; r² = 0.79 and 0.88, respectively), suggesting that both sFe and cFe contributed significantly to the TPD meteoric water dFe source. However, the slope of the sFe versus %MET regression was ~90% that of the corresponding dFe slope (cFe slope was only 47% that of dFe), suggesting that the dFe that persists from meteoric water fluxes is rich in smaller organically chelated Fe species and is not predominantly composed of colloidal-sized inorganic Fe oxyhydroxides. This makes sense since larger colloidal-sized species such as weathered Fe nanoparticles present in river water and Fe oxyhydroxides formed during Fe precipitation within the Siberian estuaries were likely removed from the water column via scavenging and/or aggregation. In contrast, smaller soluble-sized species are more likely to persist as Fe is bound to organic ligands, which have recently been found to be dominated by humic substances that stabilize Fe and account for its high solubility in the TPD (Laglera et al., 2019; Slagter et al., 2017, 2019). Other strong organic Fe-binding ligands may also contribute to the stabilized sFe pool, as Fe bound to strong organic ligands have been found to be resistant to flocculation in estuaries (Bundy et al., 2015).

Fe stable isotope ratios (δ⁵⁶Fe) support a riverine source for dFe in the TPD (Figure 5d). If reducing shelf sediments were a source of dFe to the TPD, then the dFe δ⁵⁶Fe signature would be expected to be low (−2‰ to −4‰; Conway & John, 2014; John et al., 2012; Severmann et al., 2006, 2010). Instead, TPD δ⁵⁶Fe values are much closer to continental values around 0‰ (ranging from −0.4‰ to +0.5‰ in waters with >5% meteoric water), consistent with those found in other Arctic rivers (Ilina et al., 2013; Stevenson et al., 2017; R. Zhang, John, et al., 2015). Indeed, the values are very similar to the range in δ⁵⁶Fe for particulate and colloidal Fe in the Lena River Estuary (−0.4‰ to +0.1‰), though dissolved δ⁵⁶Fe was not measured in that study (Conrad et al., 2019).

Dissolved Fe, Co, Ni, and Cu had the most statistically significant correlations with meteoric water fraction (Figures 5,6), and these metals are likely to be substantially organically complexed in seawater (Constant M.G. van den Berg, 1995; Bruland et al., 2013; Millero et al., 2009; Yang & van den Berg, 2009). These results are consistent with the high apparent residence times of Fe, Ni and Cu determined within TPD-influenced water of the GN01 study (Kadko et al., 2019) and supports the conclusion that organic chelation is required for efficient transport of riverine dissolved metals into the central Arctic. For example, the majority of the dCo in samples with the highest fraction meteoric water was strongly organically complexed (up to 90%) compared to an average of ~70% organically complexed in the remainder of the GN01 transect.

Dissolved Fe, with its apparent and measured Eurasian river end-member of 19 nmol L⁻¹ and 2,300 nmol L⁻¹ (Holmes et al., 2019), respectively, was largely scavenged/aggregated in the estuary and/or over the shelf before being transported offshore. In contrast, the apparent riverine end-members for dCu (30 nmol L⁻¹) and dNi (31 nmol L⁻¹) are within less than one order of magnitude of their Eurasian Arctic river averages (dCu = 22 nmol L⁻¹; dNi = 17 nmol L⁻¹; Holmes et al., 2019), suggesting that they were organically chelated before leaving the river and were not scavenged or taken up biologically in the estuary or on the shelf to any significant extent (C.M.G. van den Berg & Nimmo, 1987; Donat & van den Berg, 1992). The 100% meteoric water end-member yields dissolved and labile Co apparent end-members of 854 and 243 pmol L⁻¹, respectively. The dCo apparent riverine end-member is largely consistent with the Arctic rivers average of 1,280 pmol L⁻¹ (Holmes et al., 2019), suggesting that rivers are the primary source of dCo in the TPD and that this riverine dCo signature is preserved by organic matter complexation (e.g., Ellwood & van den Berg, 2001) and mainly affected by dilution over the time scale of transport from the shelf to the sampling region in the vicinity of the North Pole (~6–12 months, Kipp et al., 2018). As was observed for dNi and dCu, these data suggest that organic complexation from the high DOM riverine waters plays an important role in stabilizing scavenging-prone metals in the TPD.

Dissolved Zn is also elevated in Eurasian Arctic rivers (59 nmol L⁻¹; Holmes et al., 2019), but compared to other metals there was only a weak, but positive relationship between dZn and meteoric water to link the central Arctic Zn to a riverine source (Jensen et al., 2019). Dissolved Zn concentrations are highest in halocline waters of the western Arctic (~1–6 nmol L⁻¹; 50–250 m) due to a large source of Zn from remineralization in Chukchi Shelf porewaters (Jensen et al., 2019), while they decrease to subnanomolar concentrations in the uppermost 50 m of the Arctic under the ice. Because the concentrations of Zn in Arctic melt pond waters (Marsay et al., 2018) and the Lena River (Guieu et al., 1996) are also around 1 nmol L⁻¹, it is difficult to distinguish the source of Zn in the TPD waters from concentrations alone. Zinc stable isotopes (δ⁶⁶Zn) range between +0.2‰ and +0.8‰ in TPD waters with a significant meteoric water component (>10%), a range that is consistent both with the isotope composition of meltwaters (−0.11‰ to +0.75‰; Marsay et al., 2018) and with the variability found in global rivers (−0.12‰ to +0.88‰; Little et al., 2014). Thus, the lower δ⁶⁶Zn signatures observed in the core of the TPD compared to surface waters further south in the Arctic (Figure 5k) may reflect either background seawater variability, sea ice input, or river/shelf input from the TPD, though the positive slope between dZn and meteoric water does suggest that the latter source is most likely (Bruland, 1989; Murray et al., 2014). Further research will be needed to determine the relative importance of these sources, though it has been suggested that aerosol input is likely the most significant source of Zn to the open Arctic with some riverine input as well (Kadko et al., 2019).

Dissolved Cd concentrations in Eurasian Arctic rivers (68 pmol L⁻¹; Holmes et al., 2019) are lower than those in the central Arctic, meaning that both dCd concentrations and Cd stable isotopes (δ¹¹⁴Cd) in the TPD are heavily influenced by a marine source (Figure 5i). Additionally, δ¹¹⁴Cd in both melt ponds and Arctic rivers is similar to seawater values, making it difficult to know whether the lower δ¹¹⁴Cd values seen in the TPD are caused by river/shelf input, or are simply the result of the mixing between Atlantic and Pacific waters in the Arctic (Zhang et al., 2019).

Dissolved Al did not trend significantly with the meteoric fraction, while dGa did, but with an inverse relationship. Gallium and Al have similar oceanic sources (rivers and atmospheric deposition) and sinks (scavenging), although Ga is less particle reactive and thus has a longer residence time in seawater (Orians & Bruland, 1988; Shiller, 1998). Surface dAl and dGa concentrations are both low in the Arctic Ocean, including the TPD, suggesting that there is little or no input from shelf waters (Figures 6a and 6c). The dAl concentrations are also much lower than those typically observed in the open ocean where dust deposition can lead to surface enrichment of this TEI (e.g., the Atlantic Ocean; Figure 7c). There was no evidence of ice rafted sediment adding trace elements to the surface waters, which is consistent with the lack of any visual identification reports of this material during the GN01 cruise. Dissolved Al during the 2007 GIPY11 cruise was 0.3–1.4 nmol L⁻¹ in surface waters of the eastern Arctic (Schlitzer et al., 2018), which is very similar to the range of data reported herein (0.9–1.8 nmol L⁻¹). In contrast, the 1994 Trans Arctic cruise (Measures, 1999) showed highly variable concentrations of dissolved Al, with the highest values corresponding to regions where significant levels of ice rafted sediment were observed. Furthermore, though the fluvial dAl load in Arctic rivers is known to be substantial (Stoffyn & Mackenzie, 1982), export of riverine Al to the open ocean, including in the TPD, is negligible because of estuarine removal processes (Mackin & Aller, 1984; Tria et al., 2007). In support of the estuarine/coastal removal argument, previous data (Middag et al., 2009) showed that dAl values are as low as 2.3 nmol L⁻¹ in the Laptev Sea, which is strongly influenced by the outflow of the Lena River. While fluvial dGa data are limited, most reports suggest typical concentrations in the 10–100 pmol L⁻¹ range (Colombo et al., 2019; Gaillardet et al., 2014; Shiller & Frilot, 1996). Additionally, McAlister and Orians (2012) demonstrated that dGa behaves nonconservatively in the Columbia River plume, showing apparent scavenging removal. Likewise, McAlister and Orians (2015) indicate that interactions with Arctic shelf sediments may be a dGa sink due to scavenging onto particle surfaces, which is supported by the negative slope and 100% meteoric water intercept of −24 pmol L⁻¹ reported herein.

Nonconservative behavior is observed in the global ocean vanadium cycle as a result of biological uptake, particle scavenging, and reduction to an insoluble phase (Brinza et al., 2008; Crans et al., 2004; Wehrli & Stumm, 1989). However, V in surface water may also be diluted by sea ice melt and/or riverine input (Marsay et al., 2018). The Arctic riverine end-member ranges between 1.7–22 nmol L⁻¹; the discharge-weighted mean of Eurasian rivers is 11 nmol L⁻¹ (Holmes et al., 2019). Like dGa, the dV and meteoric water correlation in the TPD has a negative slope and is significant (r² = 0.65, p = 0.001; Figure 6b), with an extrapolated dV river end-member concentration of −35 nmol L⁻¹. This negative zero salinity intercept requires a removal process at lower salinities, likely related to coupled particle scavenging and reduction in shelf waters and sediments (Kadko et al., 2019; Morford & Emerson, 1999; Whitmore et al., 2019). Specifically, release of dissolved Fe from reducing sediments leads to precipitation of Fe oxyhydroxides in the water column which then scavenge dV (Trefry & Metz, 1989), delivering the sorbed V to the sediments with the settling Fe particles. In the reducing sediments, the Fe can then be dissolved and released to the water again, while reduced V can be more permanently incorporated into the sediments (Scholz et al., 2011). This cycling of the Fe between sediments and water column creates an “Fe shuttle” that can lower the dV of the water column (Scholz et al., 2011, 2017). As noted in section 4.1.5, the changing ²²⁸Ra distribution implies increased sediment-water exchange on the Arctic shelves (Kipp et al., 2018; Rutgers van der Loeff et al., 2018). How this impacts the Fe shuttle and the resulting effect on dV is unclear since increased sediment-water exchange could increase sediment efflux of reduced Fe or diminish it through sediment oxygenation.

The projected future increases in Arctic riverine discharge and thawing permafrost are likely to enhance fluxes of bioactive trace elements such as Fe, Co, Ni, and Cu, which are high in rivers and sediments, as well as fluxes of DOM-derived organic ligands that stabilize these metals in the dissolved phase. A future Arctic with higher concentrations of DOM might serve to further enhance some of these metal fluxes, as organic chelation appears to be critical in selecting the dissolved metals that are stably transported via the TPD (Laglera et al., 2019; Slagter et al., 2017). Thus, future increases in meteoric water in the Arctic may disproportionately impact certain trace metals over others, with implications for organisms that use those metals as cofactors in essential metabolic processes. For example, some phytoplankton use Co instead of Zn in carbonic anhydrase, despite that concentrations of Zn greatly exceed those of dCo in surface waters of most regions (Saito & Goepfert, 2008). However, the Arctic is unique in that surface Co/Zn ratios are much higher than in other ocean basins. Higher inputs of Co to the TPD may also affect biological communities downstream of the Arctic, as TPD waters exit the Arctic through Fram Strait into the North Atlantic Ocean. Evidence of this elevated Co signature has been observed in the North Atlantic (Noble et al., 2017). Increases in these metal sources in a future warming Arctic may therefore impact not only Arctic biological community structure, but potentially North Atlantic ecosystems as well.

Additionally, with more open water from ice cover decline, atmospheric deposition of Al and Ga may become more important sources of these two TEIs to the surface Arctic. However, this may be offset in part by scavenging loss due to increased shelf sediment-water interactions (Kipp et al., 2018; McAlister & Orians, 2012, 2015). The contrast in the residence times for dGa and dAl suggests that comparison of the distributions of the two elements in a changing Arctic Ocean may reveal changes in scavenging and resuspension impacts on Al and other elements. Lastly, shelf conditions that promote hypoxia may influence the V cycle by increasing V removal by the reducing sedimentary environment.

4.1.7 Mercury

Among all stations located north of 84°N and shallower than 50 m, total mercury (tHg) ranged from ~0.5–2.5 pmol L⁻¹, methyl-mercury (MeHg, the sum of monomethyl and dimethyl mercury) ranged from <0.05–0.22 pmol L⁻¹, monomethyl-mercury (MMHg) ranged from <0.05–0.20 pmol L⁻¹, and dimethyl-mercury (DMHg) ranged from <0.05–0.12 pmol L⁻¹ (Figures 6f and 6g) (Agather et al., 2019). Contrary to all other open ocean basins, total Hg concentrations were enriched in surface waters. Total Hg and MeHg correspond well to the few previous observations available in the central Arctic Ocean (Heimbürger et al., 2015) and the Canadian Arctic Archipelago (F. Wang et al., 2012; K. Wang et al., 2018): tHg surface enrichment followed by a shallow MeHg peak at the halocline and in the Atlantic waters below. Prior to the 2015 GEOTRACES campaign, there were no MMHg and DMHg data for the central Arctic Ocean. Similar to other open ocean basins, MeHg concentrations were depleted in surface waters, likely due to a combination of MeHg photodemethylation, MMHg uptake into phytoplankton and DMHg evasion to the atmosphere. Although ice can act as a barrier to air-sea gas exchange and hinder elemental Hg (Hg^o) evasion (DiMento et al., 2019), no significant differences were observed between the MeHg concentrations at ice-covered versus non-ice-covered stations. Looking forward, ice thinning and melting in the central Arctic (Krumpen et al., 2019) may reduce this barrier.

Samples with elevated meteoric water fractions (>15%) were characterized by higher tHg concentrations (up to ~2 pmol L⁻¹), though there was no significant correlation between the two variables. This might be because rivers are not the only source of tHg to the water column. Mercury also enters the Arctic Ocean via atmospheric deposition and oceanic inputs, mostly from the Atlantic Ocean (Cossa et al., 2018; Outridge et al., 2008; Soerensen et al., 2016; Sonke et al., 2018). However, the lack of correlation between total Hg and meteoric water input in the TPD is surprising given the substantial input flux predicted from measurements of Hg in Arctic rivers. Sonke et al. (2018) derived a discharge-weighted tHg concentration of 46 pmol L⁻¹ for the monitored Eurasian rivers, with values of up to 191 pmol L⁻¹ in the spring freshet (Yenisei River). This result implies a large loss of Hg in estuaries and shelves, which may be the result of atmospheric evasion (Fisher et al., 2012; Sonke & Heimbürger, 2012). A more recent box model study reveals that a portion of the evading Hg is in the form of DMHg (Soerensen et al., 2016). Estuarine and shelf sediments might also act as sinks for Hg entering from pan-Arctic rivers (e.g., Amos et al., 2014), but this idea remains to be tested for this basin.

The MeHg species had no significant correlation to meteoric water fraction above 84°N. Since shelf sediments can be sources of MeHg (e.g., Hammerschmidt & Fitzgerald, 2006; Hollweg et al., 2010), we might expect a correlation to meteoric water inputs. The lack of such a correlation suggests that either MeHg produced on the Eurasian shelves was lost to demethylation processes during the ~6–18 month transit from the shelf-break to the central Arctic Ocean, or that production in the mixed layer is a stronger source than the shelves. Large subsurface maxima in methylated Hg species (Agather et al., 2019; Heimbürger et al., 2015) suggests a third source for MeHg in the TPD could be diffusion from the MeHg species-rich halocline (Soerensen et al., 2016).

In the future, climate warming is expected to increase Hg inputs to the Arctic drastically as permafrost contains large Hg stocks (Schuster et al., 2018). The Arctic reservoir with the highest relative proportion of MeHg, often representing more than 40%, is generally open ocean seawater (Heimbürger et al., 2015). It is primarily the ocean-sourced MeHg that bioamplifies to harmful levels, putting Arctic wildlife and human health at risk. The additional input of Hg and DOC might further stimulate MeHg production in the Arctic Ocean. Future coupled ocean–atmosphere numerical models (e.g., Fisher et al., 2012; Zhang, Jacob, et al., 2015) and box model assessments (e.g., Soerensen et al., 2016) designed to constrain Arctic Hg cycling will need to consider Hg cycling and transport associated with the TPD.

4.1.8 Rare Earth Elements

North of 84°N, dNd and dEr concentrations in the upper 50 m ranged between 16.8 and 48.6 pmol L⁻¹ and between 4.5 and 13.8 pmol L⁻¹, respectively. Surface concentrations of both rare earth elements (REEs) are highest within the TPD, a factor of at least 2 higher than outside the TPD influence (Figures 6h and 6i). The spatial and vertical REE distributions seen in this new data set are in agreement with previous REE distributions from the central Arctic, including extremely high concentrations within the TPD (Figure 2h;Andersson et al., 2008 ; Porcelli et al., 2009 ; Zimmermann et al., 2009), exceeded only by REE concentrations in the Laptev Sea close to the Lena Delta (Laukert et al., 2017). The elevated surface REE concentrations in the central Arctic, and in particular within the TPD, are supportive of a substantial terrestrial input mainly via rivers. This inference is reinforced by the correlation of dNd and dEr with the fraction of meteoric water (Figures 6h and 6i; r² = 0.54 and 0.49 for Nd and Er, respectively; p < 0.001 for both) and with DOC (Figure 3d). Together with a lack of systematic changes in Er/Nd ratios in the TPD, this suggests little to no particle-seawater interactions in the central Arctic. This is in contrast to other open ocean settings, where scavenging removes REEs with a preferential removal of the light over the heavy REE and therefore the Er/Nd ratio (e.g., Elderfield, 1988).

The mean discharge weighted Eurasian river dNd concentration is 1,300 nmol L⁻¹ (Holmes et al., 2019), though there is variability among the Siberian rivers as discussed by Zimmermann et al. (2009) for the summer high discharge season (Ob: 2,152 pmol L⁻¹; Lena: 826 pmol L⁻¹; Yenisei: 154 pmol L⁻¹). This variability may in part explain some of the scatter in the REE versus meteoric water fraction correlation, especially at high values. The apparent river end-member REE concentrations that were estimated by extrapolating the REE concentrations to 100% meteoric water (Nd = 120 pmol L⁻¹, Er = 32.3 pmol L⁻¹) are therefore difficult to compare to known river REE compositions. If the weighted river Nd concentration of 1,300 pmol L⁻¹ is combined with an estuarine removal of 75% Nd (Laukert et al., 2017), the river end-member Nd concentration reaching the shelf would be ~325 pmol L⁻¹, nearly a factor of three higher than the extrapolation of the REEs to 100% meteoric water. This discrepancy would therefore imply greater estuarine REE removal, substantial REE removal over the shelf, and/or removal along the flow path of surface waters from the shelves to the central Arctic. Removal over the shelf or along the TPD is, however, not reflected in the heavy over light dissolved REE ratios (Er/Nd = 3–4) that otherwise should be elevated by scavenging due to preferential removal of the light relative to the heavy REEs (Elderfield, 1988), but are similar to those in the Laptev Sea outside the direct river influence, that is, at high salinity. Therefore, the extrapolated river end-member Nd concentration approach must be hampered by the large differences in river REE concentrations, seasonal variability in the river end-members and/or discharge and the unknown relative contributions of the rivers to the REE signal in the TPD.

Unlike the radium isotope increase that has been attributed to enhanced input from the Siberian shelves in response to decreasing sea ice cover (Kipp et al., 2018; Rutgers van der Loeff et al., 2018), there are no systematic REE changes seen in the central Arctic from 2003 to 2015 (Andersson et al., 2008; Porcelli et al., 2009; Zimmermann et al., 2009). Hence, a recent increase in REE input from the shelves or the rivers is not observed based on the limited data set currently available.

4.1.9 Thorium-232

Thorium-232, with a half-life of 14.01 Gyr, is the only primordial nuclide of thorium and is abundant in the continental crust (Rudnick & Gao, 2003). The dissolved ²³²Th concentrations in the TPD are an order of magnitude higher than those observed in deep waters at the same stations. Surface dissolved ²³²Th concentrations at Arctic TPD stations are a factor of three higher than those observed in the open North Atlantic Ocean, with similar concentrations at 500 m where the Arctic is dominated by Atlantic water (Jones, 2001) (Figure 7e).

A strong positive correlation between dissolved ²³²Th and meteoric water fraction is observed in upper ocean waters (<50 m) at all TPD stations (r² = 0.83, p < 0.001) (Figure 3k). An associated strong positive correlation is also observed in these waters between dissolved ²³²Th and DOC. These strong positive correlations taken together with low surface ²³²Th concentrations at margin stations suggest that the source of the Th-enriched TPD surface waters is Siberian rivers and not diagenetic release of Th from eastern Arctic shelf sediments. Further, similar to many of the transition metals described above, these correlations suggest that ²³²Th is complexed by organic ligands that prevent it from being scavenged during transport (Hirose & Tanoue, 2001; Slagter et al., 2017; R. Zhang, John, et al., 2015).

4.2 TPD Transport Rates and Fluxes

As discussed in the introduction, there are a number of studies that have documented and modeled the circulation pathways and seasonal to interannual dynamics of the Transpolar Drift (Ekwurzel et al., 2001; McLaughlin et al., 1996; Rigor et al., 2002; Rudels, 2015; Schlosser et al., 1994). However, to the best of our knowledge, there are no published estimates of the mass transport associated with the TPD. Fortunately, Kipp et al. (2018) reported two independent estimates of the transport time scale of the TPD in 2015 from the eastern Arctic shelf break (78°N) to ~88°N. Using radionuclide tracers (²²⁸Th/²²⁸Ra), transport was in the range of 6–12 months or 0.04–0.08 m/s. Ice back trajectory analysis suggested a time scale of 8 to 18 months for the GEOTRACES stations within the TPD; assuming a transport distance of 1,200 km, mean current speeds would be in the range of 0.03–0.06 m/s. Separately, Kadko, Galfond et al. (2016) modeled the decay of atmospherically derived ⁷Be during transport under the ice to derive an average TPD current speed of 0.10 m/s. Combined, the average current speed for the TPD is thus 0.059 ± 0.030 m/s, which is equivalent to a mass transport of 0.9 ± 0.4 Sv if we assume the width and depth of the current are 600 km and 25 m, respectively (see section 3 for details on how the horizontal extent of the TPD was defined). Considering that this current is approximately 20% meteoric water, the lower bound of this flux estimate (0.5 Sv × 0.2 = 0.10 Sv) compares well with the total Eurasian river runoff estimates of 2,800 km³/yr or 0.09 Sv (Lammers et al., 2001) and 3,264 km³/yr or 0.10 Sv (Sonke et al., 2018). This ground-truthing exercise does not account for direct precipitation to Arctic Ocean surface waters, which is of the same order of magnitude as river runoff on a basin-wide scale (Serreze et al., 2006). However, the fact that the Eurasian river runoff is more closely aligned with the lower bound of our meteoric water-corrected mass transport estimate may also be owed to (1) an overestimate in our TPD cross sectional area or (2) a larger contribution of North American rivers to the freshwater component of the TPD. This latter point might be a further explanation for the scatter in some of our TEI/meteoric water relationships.

Fluxes of TEIs with strong linear correlations to meteoric water can be reasonably estimated because we can assume they are approximately conserved away from the shelves. These include dFe, dCu, dNi, d²³²Th, dNd, and DOC. For the purposes of these flux calculations, we assume that the TPD source waters leaving the Siberian shelves consist of 20% meteoric water. Dissolved Fe (r² = 0.67) had a concentration equal to 3.6 nmol L⁻¹ at 20% meteoric water. With a mass transport of 0.9 Sv, the shelf derived Fe flux in the TPD is 1.0 ± 0.5 × 10⁸ mol yr⁻¹ (Table 3). Kadko et al. (2019) quantified the atmospheric deposition of soluble Fe to the GEOTRACES TPD stations as 0.61 ± 0.17 nmol/m²/d. Scaling this estimate up to the Arctic Ocean basin (9.5 × 10¹² m²; Jakobsson, 2002) produces a total atmospheric deposition flux of 2.1 ± 0.6 × 10⁶ mol yr⁻¹, which is only 2% of the shelf-basin flux of Fe derived from the TPD. Kadko et al. (2019), using an approach that is independent to our own, arrived at a similar conclusion that the atmospheric input is dwarfed by that delivered through the TPD. By comparison, the shelf-basin flux of dissolved Fe from the Chukchi Sea has been estimated at 9.1–22 × 10⁶ mol yr⁻¹ (Vieira et al., 2019); hence, the Siberian shelves and the TPD are much more efficient transporters of continental margin-derived dissolved Fe to the central Arctic basins than dust. A similar picture emerges for both Cu (2.0 ± 1.0 × 10⁸ mol yr⁻¹; Table 3), for which atmospheric deposition is only 4% of the shelf input, and Ni (2.4 ± 1.2 × 10⁸ mol yr⁻¹; Table 3), with a TPD flux that rivals global riverine Ni inputs (3.6 × 10⁸ mol yr⁻¹; Cameron & Vance, 2014). These disparities highlight the relative importance of boundary TEI inputs in this shelf-dominated and relatively small ocean basin.

Table 3. Fluxes to the Central Arctic Ocean in Association With the Transpolar Drift

Dissolved species	Concentration^a	Units	Flux (mol/yr)	Error
Fe	3.6	nmol L^−1	1.0E+08	5.1E+07
Cu	7.0	nmol L^−1	2.0E+08	9.9E+07
Ni	8.6	nmol L^−1	2.4E+08	1.2E+08
^232Th	1.3	pmol L^−1	3.5E+04	1.8E+04
DOC	132	μmol L^−1	3.7E+12	1.9E+12
Nd	44	pmol L^−1	1.2E+06	6.2E+05

• a At 20% meteoric water.

Given the highly particle reactive nature of thorium, the dissolved ²³²Th concentrations observed within the TPD were somewhat surprising. In terms of flux, the TPD is estimated to carry 3.5 ± 1.8 × 10⁴ mol yr⁻¹ of ²³²Th. If the dust input and solubility of ²³²Th are ~50 μg/m²/yr (Kienast et al., 2016) and 1% (Hsieh et al., 2011), respectively, then the dissolved input to the Arctic Ocean (assuming no ice cover) would be on the order of 2.1 × 10⁴ mol yr⁻¹. Such a strong lateral input of ²³²Th relative to atmospheric deposition would complicate the use of dissolved ²³²Th as a dust flux proxy in this basin and potentially others where fluvial inputs rich in DOM are entering the ocean (Anderson et al., 2016; Hayes et al., 2013; Kienast et al., 2016; Robinson et al., 2008).

The riverine input of Nd to the Arctic Ocean has not been well constrained owing to the relatively few measurements of Nd concentrations in eastern Arctic rivers and of Nd removal in the estuarine mixing zone. That said, the TPD flux of Nd is 1.2 ± 0.6 × 10⁶ mol yr⁻¹, which is on par with the global riverine Nd flux (1.8 × 10⁶ mol yr⁻¹; Arsouze et al., 2009). From recent work we know that sedimentary fluxes of Nd to the ocean may far exceed those from rivers (18–110 × 10⁶ mol yr⁻¹; Abbott et al., 2015). Yet, the Nd carried in the TPD is lower than expected from a combined discharge-weighted river contribution and hence rather suggests a Nd deficit, providing no evidence for Nd contributions from shelf sediments in the Arctic. Rather, the lower than expected REE concentrations point to the large range in river end-member REE concentrations, the likelihood of substantial REE removal in the estuaries, and their unknown relative contributions to the TPD. For DOC, the flux carried by the TPD is 3.7 ± 1.9 × 10¹² mol yr⁻¹, as compared to a total DOC flux from the major Arctic rivers of 2.1–3.0 × 10¹² mol yr⁻¹ (Raymond et al., 2007) indicating a major contribution of terrigenous DOC to the total TPD DOM flux.

5 Conclusions

Intensification of the hydrologic cycle and permafrost degradation may result in the release of about 25% of the carbon stored in Arctic soils in the next 100 years (Gruber et al., 2004). According to the National Oceanic and Atmospheric Administration Arctic report card (Osborne et al., 2018), the 2018 summer/autumn discharge for the largest rivers flowing into the Arctic was 20% greater than in the 1980–1989 period and will continue to increase. These changes will have a substantial effect on the riverine supply of DOM into the Arctic Ocean, as well as the long-distance transport of TEIs within the Transpolar Drift that are likely complexed by this organic matter, including Fe, Co, Ni, Cu, Th, and possibly the REEs.

While the halocline contains ample nutrient concentrations, the increased freshwater inputs are strengthening water column stratification, which could further limit nutrient inputs via vertical mixing processes (Rudels et al., 1991). Hence, increased macronutrient and micronutrient concentrations delivered to the central Arctic Ocean via the TPD may play an important role in upper ocean productivity in the coming decades, since, for example, nitrate already limits primary production in some Arctic locations (Tremblay & Gagnon, 2009), as did Fe in the case of one large under ice bloom, though this was located outside of the TPD (Rijkenberg et al., 2018). In the case of Fe, whether limitation will occur in the changing Arctic will depend on the interplay between nutrient utilization ratios (Rijkenberg et al., 2018) and projected increases of ligand-borne, specifically humic-borne, terrestrial dFe (Slagter et al., 2017, 2019).

The complexity of physical and biochemical factors and their interplay, such as the effect of increased river runoff and stratification on the saturation state of aragonite (Yamamoto-Kawai et al., 2009), combine with scarcity of data to make future effects of TPD influence on the central Arctic difficult to predict (Carmack & McLaughlin, 2011). However, DOM is strongly related to hydrographic parameters and biogeochemical cycles in the shelf seas and TPD (Amon et al., 2003; Granskog et al., 2012), but has the advantage of relatively simple measurement via remote sensing in ice-free waters (Fichot et al., 2013; Juhls et al., 2019; Matsuoka et al., 2017) or in situ instrumentation capable of high vertical resolution such as the fluorometers deployed on these cruises. Looking to the future, this makes CDOM a powerful tracer of climate change impacts on a multitude of Arctic system processes (e.g., Stedmon et al., 2015).

For some TEIs, the sediments within the broad and shallow eastern Arctic shelves play a dominant role in their cycling and signature within the TPD. Radium isotopic ratios and a mass balance calculation point to shelf sediments as the dominant source of ²²⁸Ra carried by the TPD (Kipp et al., 2018). While ²²⁸Ra is not a biologically important TEI, it acts as a quasi-conservative tracer of other shelf-derived materials like Ba, which has TPD concentrations that cannot be fully explained by a river source. The Ra-derived evidence of active sediment-water exchange processes in the eastern Arctic coastal zone supports the apparent strong sinks for Pb and V, which are known to be removed by particle scavenging and/or reduction processes in shelf sediments. Increased ²²⁸Ra levels in the TPD therefore suggest that the concentrations of these other TEIs may be affected both positively (Ba) or negatively (Pb, Al, Ga, and V) under a changing climate where shelf sediments are exposed to wind-driven mixing under reduced ice cover. This ice loss and its potential impacts on TEI cycling will be exacerbated not only by atmospheric forcing, but also by penetration of warmer Atlantic Ocean waters into the Arctic (e.g., Polyakov et al., 2017).

The TEIs that have the strongest correlation with meteoric water fraction are those that are known to form complexes with organic matter. As a result, other than dilution via mixing, their concentrations, which are significantly elevated relative to other ocean basins, are preserved in the TPD over distances >1,000 km and timescales of up to 18 months. It is therefore reasonable to expect that this TEI “fingerprint” of the TPD would be carried beyond the ice covered central Arctic Ocean, through Fram Strait, and into the ice-free surface waters of the North Atlantic Ocean as seen for Arctic river DOM (Amon et al., 2003; Benner et al., 2005; Gerringa et al., 2015; Granskog et al., 2012). In the present day, the TEIs transported in the TPD may become participants in biogeochemical processes of this ocean basin, or in the future be utilized closer to their source as the pan-Arctic ice cover is reduced with warming temperatures. This new utilization would apply to an increasingly ice-free Arctic Ocean including the Canada Basin, where the Beaufort Gyre (Figure 1) is known to entrain and store an increasing amount of freshwater sourced from eastern Arctic rivers (Giles et al., 2012; Morison et al., 2012; Rabe et al., 2011, 2014).

Lastly, our understanding of the effects of the changing climate on Arctic Ocean TEI concentrations and fluxes has been greatly hampered by a lack of data, mainly due to the logistics and expense of conducting oceanography at high-latitudes where icebreakers are required for sampling. Geopolitical issues have resulted in large data gaps for the eastern Arctic shelf seas. In the near future, international collaboration through long term observatories at key locations and Arctic gateways, synoptic surveys (e.g., http://www.synopticarcticsurvey.info) and advances in technology (e.g., floats, gliders, ice tethered sensors and samplers) may provide the temporal and spatial coverage needed to address some of the pressing unanswered questions posed herein.

Acknowledgments

The authors thank the Editor and two anonymous reviewers for their constructive comments and feedback. This study would not have been possible without the dedication of the captains and crews of the USCGC Healy and R/V Polarstern. The authors also thank the many members of the shipboard scientific parties, including the chief scientists, who enabled sample collection via the various CTD rosette and in situ pumping systems, as well as the sample analyses on the ship and back in their home laboratories. Funding for Arctic GEOTRACES was provided by the U.S. National Science Foundation, Swedish Research Council Formas, French Agence Nationale de la Recherche and LabexMER, Netherlands Organization for Scientific Research, and Independent Research Fund Denmark. Data from GEOTRACES cruises GN01 (HLY1502) and GN04 (PS94) have been archived at the Biological and Chemical Oceanography Data Management Office (Biological and Chemical Oceanography Data Management Office (BCO-DMO); https://www.bco-dmo.org/deployment/638807) and PANGAEA (https://www.pangaea.de/?q=PS94&f.campaign%5B%5D=PS94) websites, respectively. The inorganic carbon data are available at the NOAA Ocean Carbon Data System (OCADS; doi:10.3334/CDIAC/OTG.CLIVAR_ARC01_33HQ20150809). In addition to these national data archives, the data used in this paper is available as supporting information.

Appendix A: Methods Appendix

The methods for TEIs already published can be found in the papers listed in Table A1 below. For the determination of DIC and TA concentrations, methods are previously published in Ulfsbo et al. (2018) and Woosley et al. (2017). However, the joint description of the two cruises and the uncertainty analysis was not previously discussed. The precision from replicate sample analyses was estimated to be better than 2 μmol L⁻¹ for both DIC and TA during GN04 (Jones & Ulfsbo, 2017) and better than 2.5 and 2 μmol L⁻¹ for DIC and TA, respectively, during GN01 (Kadko, Millero et al., 2016). The accuracy of DIC and TA was determined by routine analyses of Certified Reference Materials (CRM) provided by A.G. Dickson, Scripps Institution of Oceanography, USA. Deep water (>2,000 m) mean concentrations for both cruises at the crossover station (GN04-101 and GN01-30) agree within the uncertainty of the measurements. Uncertainty propagation for the carbonate system variables was performed according to Orr et al. (2018). Based on input variables from Table 1 of Orr et al. (2018) with two of the carbonate system parameters, the combined relative uncertainty for the third measured parameter was calculated using standard uncertainties in the dissociation constants and total boron to salinity ratio given by Orr et al. (2018). The propagated combined standard uncertainty was 3.3 μmol L⁻¹, 3.2 μmol L⁻¹, and 0.012 units for DIC, TA, and pH, respectively.

Table A1. References From Which Data Were Drawn for Use in This TPD Synthesis Paper or From Which Methods Were Used to Generate Previously Unpublished Data

TEI	Reference(s)
Hydrography, inorganic nutrients (NO3^−, PO4^3−, SiO4)	Cutter et al. (2014)
δ^18O	Pasqualini et al. (2017)
Total alkalinity, dissolved inorganic carbon	Dickson et al. (2007); Kadko, Millero, et al. (2016)
DOC	Hansell (2005)
Ra isotopes	Kipp et al. (2018, 2019); Rutgers van der Loeff et al. (2018)
dBa	Ho et al. (2019); Jacquet et al. (2005)
POC, bSi, δ^13C-POC	Lam et al. (2018)
pFe, pAl, pmol L^−1n, pBa	Lam et al. (2018); Planquette and Sherrell (2012)
dFe (GN01 is average of Fitzsimmons, John and Hatta; PS94 is average of Gerringa ICP-MS and FIA analysis)	Conway et al. (2013); Hatta et al. (2015); Jensen et al. (2019); Middag et al. (2015); Rijkenberg et al. (2014)
dMn (GN01 is average of Fitzsimmons and Hatta)	Hatta et al. (2015); Jensen et al. (2019)
dCd, dZn (GN01 is average of Fitzsimmons and John)	Conway et al. (2013); Jensen et al. (2019)
dNi, dCu, dPb	Conway et al. (2013); Jensen et al. (2019); Middag et al. (2015)
Soluble and colloidal Fe ultrafiltration	Fitzsimmons and Boyle (2014); Hatta et al. (2015)
Fe-binding organic ligands	Slagter et al. (2017)
dAl	Measures et al. (2015)
dCo, LCo	Hawco et al. (2016)
δ^56Fe, δ^66Zn, δ^114Cd	Conway et al. (2013)
dGa	Ho et al. (2019)
dV	Whitmore et al. (2019)
Hg species	Agather et al. (2019); Heimbürger et al. (2015); Lamborg et al. (2012)
dNd, dEr	Behrens et al. (2016)
^232Th	Anderson et al. (2012)

• Note. Refer to the publication for the sample handling and analytical methods pertaining to the TEI of interest.

DOC was determined by high temperature combustion on a Shimadzu TOC-L, as described by Dickson et al. (2007). Consensus Reference Waters were analyzed as a regular check on the quality and reliability of the measurements (Hansell, 2005). The DOC measurements from both cruises (GN01 and GN04) were done in the same lab and were consistent at the crossover stations.

Fluorescent components of chromophoric DOM (CDOM) were measured in situ using two Dr. Haardt fluorometers with backscatter fluorescence sensors that excite at 350–460 nm and emit at 550 nm. The light source and settings in the two sensors were slightly different on the GN01 and GN04 cruises, so the two fluorometers were intercalibrated using a polynomial fit of the data collected at a crossover station to ensure comparability. The correlation coefficient for the polynomial function was r² = 0.994.

As noted in section 3, certain TEI data (dFe, dZn, dCd, dMn; Table A1) are presented as the average of data collected from multiple laboratories and different analytical methods (e.g., flow injection analysis versus inductively coupled plasma mass spectrometry); therefore, the data for these TEIs used in this paper cannot be found on any of the online databases. Please refer to the provided supporting information for the average values as well as the original data sources used to create the averages.

There was good agreement between the ²²⁸Ra and ²²⁶Ra measurements made on the GN01 and GN04 transects. At the crossover station, ²²⁸Ra activities measured on GN04 were 14.9 ± 1.0 dpm 100 L⁻¹ at 10 m and those measured on GN01 were 16.2 ± 0.6 dpm 100 L⁻¹ and 17.2 ± 0.6 dpm 100 L⁻¹ at 2 and 24 m, respectively. The ²²⁶Ra activities measured at 10 m on GN04 were 9.1 ± 0.5 dpm 100 L⁻¹, while those measured on GN01 were 11.2 ± 0.2 dpm 100 L⁻¹ and 10.9 ± 0.4 dpm 100 L⁻¹ at 2 and 24 m, respectively.

The REE data collected at the crossover station (stations HLY1502–30 and PS94–101) within 14 days compare well especially below 500 m water depth, where they are within 1 pmol L⁻¹ for Nd and 0.5 pmol L⁻¹ for Er. Given the typical analytical uncertainties of ±0.4 pmol L⁻¹ for Nd and ± 0.2 pmol L⁻¹ for Er (GN04), the agreement between the two data sets is excellent. Moreover, both laboratories participated in a small interlaboratory comparison of dissolved REE in seawater aliquots from 3,000 m water depth at SAFe North Pacific station. The results of all four participating labs agreed within 3.3% RSD for all REE except Ce, Gd, and Lu (Behrens et al., 2016).

The GN01 and GN04 Hg species data were intercalibrated at the crossover station. Total Hg concentrations agreed well for the majority of sampling depths. The tHg and MeHg data have been intercalibrated and validated by the GEOTRACES standards & intercalibration committee.

The GN01 and GN04 expeditions collected and analyzed thorium samples from the crossover station in the Makarov Basin and are well-intercalibrated with the exception of a few outliers, which are likely due to the uncertainty in ²³²Th blank corrections in the GN04 data set. Th analyses were shared among four groups: LDEO and U. Minnesota for GN01, AWI (Valk et al., 2019) and LSCE (Gdaniec et al., 2020) for GN04. These four groups had participated in the GEOTRACES intercalibration exercise on Th isotopes (R. F. Anderson et al., 2012). A new intercalibration paper based on this Arctic work and focusing in particular on shallow samples is in preparation.

Erratum

In the originally published version of this article, the average discharge weighted silicate concentration for the Eurasian rivers reported by the A-GRO was incorrectly given as 177 μmol L⁻¹. The correct value is 83 μmol L⁻¹. The affected sentences on page 7 of the article have been corrected, and this may be considered the official version of record.

In the originally published version of this article, columns 4 and 5 of row 5 of Table 3 were incorrectly given as 3.7E+15 and 1.9E+15, respectively, due to a conversion error in Microsoft Excel. The values presented in the main text of the article were correct. This has been corrected, and this may be considered the official version of record.