Changes in Circulation and Particle Scavenging in the Amerasian Basin of the Arctic Ocean over the Last Three Decades Inferred from the Water Column Distribution of Geochemical Tracers

Abstract Since the 1980–1990s, international research efforts have augmented our knowledge of the physical and chemical properties of the Arctic Ocean water masses, and recent studies have documented changes. Understanding the processes responsible for these changes is necessary to be able to forecast the local and global consequences of these property evolutions on climate. The present work investigates the distributions of geochemical tracers of particle fluxes and circulation in the Amerasian Basin and their temporal evolution over the last three decades (from stations visited between 1983 and 2015). Profiles of 230‐thorium (230Th) and 231‐protactinium (231Pa) concentrations and neodymium isotopes (expressed as εNd) measured in the Amerasian Basin prior to 2000 are compared to a new, post‐2000s data set. The comparison shows a large scale decrease in dissolved 230Th and 231Pa concentrations, suggesting intensification of scavenging by particle flux, especially in coastal areas. Higher productivity and sediment resuspension from the shelves appear responsible for the concentration decrease along the margins. In the basin interior, increased lateral exchanges with the boundary circulation also contribute to the decrease in concentration. This study illustrates how dissolved 230Th and 231Pa, with εNd support, can provide unique insights not only into changes in particle flux but also into the evolution of ocean circulation and mixing.


Introduction
The pathways that waters follow in the ocean and the physicochemical changes they undergo along these pathways dictate the redistribution of key parameters such as heat, salt, energy, nutrients, and pollutants.
It is of particular interest to examine the pathways and fate of waters in the Arctic Ocean because some of these waters eventually exit to enter the Atlantic Ocean and impact North Atlantic deep water formation and the global overturning circulation (Aagaard et al., 1985;Hansen & Østerhus, 2000). Arctic Ocean waters are made of Pacific and Atlantic Ocean waters that enter the Amerasian Basin through the shallow Bering Strait and the Eurasian Basin through the shallow Barents Sea and deep Fram Strait. They eventually return to the Atlantic Ocean through Fram Strait and the relatively shallow Canadian Arctic Archipelago as upper or intermediate waters (Aagaard & Greisman, 1975). During their transit in the Arctic Ocean, these waters are transported by the upper wind-driven circulation and the intermediate cyclonic topographical-steered circulation (Figure 1; Rudels, 2001;Rudels et al., 2013).
Investigating the pathways and fate of waters in the Arctic Ocean is all the more important as the Arctic Ocean represents one of the most rapidly changing regions of the world's oceans (Pörtner et al., 2015). With a reduction in seasonal sea ice coverage, changes in hydrography and water mass circulation have been observed, not only at the surface but also at depth (e.g., Polyakov et al., 2013). Particle concentrations are also projected to increase in the rapidly changing Arctic Ocean, as a result of increased continental runoff and biological production, impacting the chemical properties of Arctic waters and downstream (e.g., Carmack et al., 2006).
The first observed changes in temperature-abnormally warm Atlantic water (hereafter, awAW)occurred in 1990 in Fram Strait (Quadfasel et al., 1991). This awAW propagated through the whole Arctic, first into the Eurasian Basin, then beyond the Lomonosov Ridge, and into the Amerasian Basin. This temperature anomaly was observed in the southern Makarov Basin (MB) and Mendeleyev Ridge in 1993 , in the central MB in 2000 (Kikuchi et al., 2005), at the northern tip of the Northwind Ridge in 2003 (Woodgate et al., 2007), and within most of the Canada Basin (CB) in 2007(McLaughlin et al., 2009; see Figure 1 for the basin and ridge locations). However, these studies did not reveal significant changes in circulation related to the awAW propagation in the Arctic. Likewise, no significant changes in particle distribution were reported from limited measurements in the CB conducted between 2003 and 2008 (Jackson et al., 2010). However, the latter study documented a sharp contrast between low particle concentrations of the central basin and much higher concentrations near the CB continental margin.
Because of their affinity for particles and residence times similar to the time scale of regional changes occurring in the Arctic, geochemical tracers such as 230 Th and 231 Pa are powerful tools to quantify the recent evolution of particle flux and circulation in the Arctic Ocean. 230 Th and 231 Pa are produced in the ocean by radioactive α-decay of uranium-234 ( 234 U) and uranium-235 ( 235 U), respectively; uranium input to the ocean occurs through continental weathering. Uranium is soluble in seawater and has a long residence time in the ocean (~400,000 years; Brewer, 1975;Chen et al., 1986), such that its concentration and the rate of production of 230 Th and 231 Pa are uniform and well known (Ku et al., 1977;Turekian & Chan, 1971). Unlike parent U, Th and Pa-especially Th-are highly insoluble in seawater and effectively removed to the sediment by adsorption onto sinking particles (Anderson et al., 1983), a process referred to as particulate scavenging. As a result, 230 Th residence times in seawater range from a few years in shallow water to a few decades in deep water; for 231 Pa, residence times range from a few decades in shallow water to a few centuries in deep water (Henderson & Anderson, 2003). If transport of 230 Th and 231 Pa by advection and turbulent diffusion can be neglected, the oceanic distribution of these radionuclides is largely controlled by reversible scavenging: dissolved radionuclides (R d ; unit: concentration) are produced continuously at fixed rates (α; unit: concentration per time) and adsorbed reversibly onto sinking particles (k a and k d are adsorption and desorption coefficients; unit: per time) to produce sinking particulate radionuclides (R p ; unit: concentration), removed at a sinking rate S (unit: length per time) to the underlying sediment (Bacon & Anderson, 1982;Nozaki et al., 1981). Thus, neglecting advection and diffusion, the conservation equations for the radionuclide dissolved and particulate concentrations ([R d ] and [R p ], respectively) are given by where z is depth, increasing downward. The sinking flux is the last term in equation (2). Under assumptions of steady state and constant S, k a and k d , equations (1) and (2) predict a linear increase of 230 Th and 231 Pa concentrations with depth, with a slope inversely proportional to the sinking rate of particles: Based on observations of a strong positive correlation between k a and the concentration of suspended matter (Bacon & Anderson, 1982), we expect to find lower concentrations and a reduced downward increase of dissolved concentrations in areas of high particle concentrations and flux. For example, this scenario is expected for the seasonally ice-free CB continental margin (Figure 2a), compared to higher concentrations and a faster downward increase in areas of low particle concentrations and flux, such as the permanently ice-covered central basin ( Figure 2c).
These simplified schematics actually reproduce quite well the 230 Th and 231 Pa oceanic distributions of the Amerasian Basin reported from the first observations, in the 1980-1990s. Low and linear vertical profiles were observed in 1995 in the southern CB, reflecting the integrated impact of particle scavenging in the boundary currents (Edmonds et al., 1998). In contrast, higher concentrations were found over the Alpha Ridge in 1983 and in the northern MB in 1991 (Bacon et al., 1989;Scholten et al., 1995; see Figure 1 for station locations). Such high concentrations of 230 Th and 231 Pa in seawater not only suggest that particle concentrations and fluxes were locally very low but also that these waters had been isolated for a few decades from the dynamic boundary circulation, allowing for in-growth of 230 Th and 231 Pa. A subsequent increase in particle flux should lead to a decrease in dissolved 230 Th and 231 Pa concentrations. Furthermore, as the affinity of 230 Th and 231 Pa changes with their particle composition (e.g., Chase et al., 2002), differences in the evolution of the distribution of 230 Th relative to 231 Pa (i.e., decoupling) could also occur depending on particle   Bluhm et al., 2015;Rudels et al., 2013). CP = Chukchi Plateau; NR = Northwind Ridge. This map and the following were created with the software Ocean Data View (Schlitzer, 2015), using the IBCAO bathymetry (version 3; Jakobsson et al., 2012).

10.1029/2019JC015265
Journal of Geophysical Research: Oceans composition (e.g., biogenic silica vs. lithogenic particles). In addition, changes in horizontal mixing or advection rates between the basin interior and the margins could decrease the contrast in 230 Th and 231 Pa concentrations between the two regions or even produce subsurface maxima in the concentration profiles (see Figure 2b). Note that the assumption of constant S and k a relies on a constant vertical flux of particulate 230 Th, which is a good approximation as organic matter decomposition does not release significant amounts of 230 Th-or 231 Pa-to seawater because of the high particle reactivity of these radionuclides.
The present work aims to investigate changes in particle scavenging and circulation over the last three decades in the Amerasian Basin by comparing 230 Th and 231 Pa concentration profiles collected in the 1980s and 1990s to 13 water-column profiles collected between 2005 and 2015. A companion paper (Yu et al., n.d.) further exploits these data by introducing 230 Th scavenging into a three-dimensional model hindcasting Arctic circulation and particle scavenging between 2002 and 2015.
We also report three dissolved profiles of neodymium isotopic compositions from 2015 that are compared to early-2000s data (Porcelli et al., 2009) and support the interpretation of the 230 Th-231 Pa profiles. Neodymium (Nd) is a rare earth element (REE) that is supplied to the ocean via continental weathering. As the neodymium isotopic composition (expressed as ε Nd ) of rocks varies with age and lithology, the ε Nd of outcropping rocks is heterogeneous (Jeandel et al., 2007). Thus, the ε Nd signal of the water masses is set by the continental Nd they receive along their circulation pathways (through direct input or through mixing with a water mass of different ε Nd ; e.g., Frank, 2002;Goldstein & Hemming, 2003;Grenier et al., 2013;Piepgras et al., 1979;Tachikawa et al., 2003). Pacific waters in the Bering Strait have ε Nd values of~−5 (Dahlqvist et al., 2007), while Atlantic waters have values of~−11 (Andersson et al., 2008). These distinct signatures allowed Porcelli et al. (2009) to trace distinct water masses and follow the evolution of their ε Nd during their transit in the Arctic in the early-2000s.

Water masses of the Amerasian Basin
The water column in the Amerasian Basin can be separated vertically into three layers, recognizable in Figures 3 and 4 and supporting information Figure S1: (i) the low-salinity Polar Surface Water (PSW) including the Polar Mixed Layer (PML) and the halocline; (ii) the warm Atlantic Water, identified as a subsurface layer with a temperature maximum bounded above and below by the 0°C isotherm, and the underlying intermediate water; and (iii) colder and more saline deep and bottom waters. The θ-S profiles presented in Figure S1 are subdivided following these three layers.  (3), that is, neglecting advection and turbulent diffusion and assuming steady state, in (a) a seasonally ice-free area, where the surface value is lower and the increase with depth is weaker due to higher scavenging and particulate sinking rates than in (c) a permanently ice-covered area. Profiles with deviations from linearity are expected in (b) areas where lateral advection of 230 Th from (a) (low grey dotted line) and (c) (high grey dotted line) occurs (grey arrows).
The PML is the homogenized surface layer undergoing brine rejection and haline convection in winter and receiving fresh water from sea ice melt in summer. The underlying halocline water originates from the Pacific Ocean and includes the fresher and warmer Alaskan Coastal Water (ACW) overlying the saltier and colder Bering Sea Water (BSW; Coachman & Barnes, 1961;Steele et al., 2004;Timmermans et al., 2014). The ACW is characterized by a subsurface temperature maximum in the salinity range [29-32.2], usually found between 50 and 100 m depth. The BSW has a summer and winter component. The Bering Sea Winter Water (BSWW) is more easily distinguished than the Bering Sea Summer Water (BSSW) and     and found between 250 and 1,000 m depth in the CB, between 250 and 800/900 m depth on the Alpha Ridge and between 200 and 800 m depth in the MB. Deeper, bounded by the 0°C temperature above and by the deep temperature minimum at about 2,400 m depth is the upper Polar Deep Water (uPDW; Rudels, 2009). Finally, below the 2,400 m temperature minimum, which marks the depth of lateral exchange between the Canada and Makarov Basins (E. Carmack et al., 2012), are the Canada Basin Deep Water (CBDW) and the Makarov Basin Deep Water (MBDW). The higher salinity of the CBDW and the MBDW compared to the Eurasian Basin bottom water could be explained by their relative isolation and the formation of dense water by brine rejection on the shelves (Aagaard et al., 1985;Jones et al., 1995;Mauritzen et al., 2013). CBDW temperature is also slightly higher, suggesting geothermal heating and mixing (E. Carmack et al., 2012), while MBDW gradually cools toward the bottom. There may be some exchange of deep water with the Eurasian Basin where the Lomonosov Ridge deepens to 1,800 m (close to the North Pole, around 175°E; Cochran et al., 2006), but the flow appears to be mostly from the MB to the Amundsen Basin (Rudels, 2009). While surface and halocline water circulation is anticyclonic in the CB and driven by the Transpolar Drift above the Lomonosov Ridge, the Atlantic layer below mainly follows topographically steered cyclonic gyres (circulation schematics in Figure 1). The circulation of bottom water remains virtually unknown. The different water masses described here have been sampled at most of the stations reported in this study (samples are identified by markers in Figures 3, 4, and S1).

Methods
New 230 Th, 231 Pa, and ε Nd data presented in this study (Tables 1 and 2 Th and 231 Pa were determined following the method described by Choi et al. (2001). For the 2015 samples the method was slightly modified as follows: (1) environmental grade acids were used to clean the sampling material and for the chemistry, to limit contamination for the determination of ε Nd on the same samples; (2) the filtration was done through a single-use 0.45 μm pore size filter cartridge (AquaPrep®; the 2011 samples remained unfiltered); (3) before analysis, the purified fractions of Th and Pa were both treated with 4 ml of concentrated perchloric acid to eliminate any organic component and then treated with concentrated HNO 3 to remove the perchloric acid. The CB samples were measured at the University of British Columbia (Vancouver, Canada) on an inductively coupled plasma mass spectrometer (Element2, Thermo Scientific) coupled to a desolvating nebulizer system (Aridus II™, Teledyne CETAC) to increase the sensitivity. 230 Th procedural blank contribution, mainly coming from the 229 Th spike, was on average 14% for the 2009-2011 samples (10.21 ± 0.87 fg) and 2% for the 2015 samples (0.99 ± 0.09 fg). The Th recovery of the two chromatographic columns was estimated for the 2015 samples and found to be 77% in average. Isotope fractionation during ICPMS analysis was estimated at 0.6% per amu. As the U/Th chromatographic separation of the 2009 samples was done 8 years after the sample collection, measured 230 Th d concentrations were corrected from a substantial 230 Th ingrowth. 231 Pa procedural blank contribution, mainly coming from the 233 Pa spike, was on average 56% for the 2009-2011 samples (46.5 ± 1.9 fg) and 70% for the 2015 samples (26.7 ± 1.6 fg). This large contribution resulted from over-spiking but was well monitored by the set of chemical blanks and accurately corrected. The Pa recovery of the two chromatographic columns was estimated for the 2015 samples and found to be 80% in average.

Nd Isotopic Composition
The REE measurements came from the same initial natural samples used for Pa and Th. They were eluted with the Th fraction during the first chromatographic column and separated from Th on the second column in the 24 ml of 8 N HNO 3 . This fraction was dried and dissolved in 1.5 ml of HCl 1 N and was processed as described in Grenier et al. (2013). After the chemical extraction and purification of Nd, samples were dissolved in 2 μl of 2 N HCl, loaded on a rhenium filament and analyzed by thermal ionization mass spectrometry (TIMS) in static mode (ThermoFisher Scientific mass spectrometer Triton, Observatoire Midi-Pyrénées,

Journal of Geophysical Research: Oceans
Toulouse). Thirteen analyses of Rennes standard and seven of La Jolla standard were performed to monitor

Hydrographic Change in the Atlantic and Bottom Waters
Comparing temperature profiles taken in the same area at different time clearly shows that the Atlantic water has generally warmed in the Amerasian Basin between the~1990s and 2010s. This warming was mainly observed in the upper part of the Atlantic water, around 300 m, for the MB and Mendeleev Ridge area, with a warming of~0.4°C between 1991 (station 176) and 2005 (station 18A) in the western MB and of~0.1°C between 1994 (station 25) and 2005 (station 11) on the Mendeleev Ridge (Figures 4a-4c). In contrast, on Alpha Ridge, the warming between 1983 (station CESAR) and 2007 (station 342) was not observed in the core of the Atlantic water layer (~450 m) but mainly between 500 and 1,500 m, in the lower part of the Atlantic layer and in the underlying uPDW. In the CB, the warming was centered in the upper part of the Atlantic water (Figures 3a-3e). An increase of temperature of~0.  Figure 5d and Tables 1 and 2). In the MB, the deep waters were warmer, fresher, and lighter in the mid-2000s than in 1991, and even slightly lighter in 2015, while the bottom MBDW were getting slightly saltier, warmer, and denser ( Figure 5h and Table 2).  Figures 6 and 7). Most of the data refer to the dissolved fraction, from filtered samples, but some represent the total fraction, from unfiltered samples. We distinguished data referring to the total fraction by reporting them in italic in Table 2. To facilitate the comparison between unfiltered and filtered profiles, dissolved profiles of 230 Th and 231 Pa were estimated at stations where samples were unfiltered, by multiplying the measured total concentrations by 0.8 for 230 Th and by 0.95 for 231 Pa (from the average dissolved/total ratios of the 230 Th and 231 Pa water column database http://climotope.earth.ox.ac.uk/data_compilations, by considering only Arctic stations, i.e., for station latitudes greater or equal to 75°N). These estimated dissolved profiles are represented by dashed lines in Figures 6 and 7. In order to track fractionation between 230 Th and 231 Pa, dissolved 231 Pa/ 230 Th activity ratios are also shown, for each area, in Figure 8.

Geochemical Tracer Results
The geochemical tracer results reported from the different areas of the Amerasian Basin compare well with the published data (Figures 6 and 7). Concentrations of 230 Th and 231 Pa are consistently higher in the MB and over ridges than in the CB. Profiles from the deep central CB and MB show smaller increase with depth in the bottom layer, reflecting particulate scavenging by sediment resuspension (Anderson et al., 1983;Rempfer et al., 2017;Scholten et al., 1995). 230 Th exhibits a greater spatial heterogeneity than 231 Pa, also consistent with the previously published profiles. Lower 230 Th concentrations are found in the sampling sites close to the boundary circulation, while higher concentrations characterize the central basins, consistent with the gradient described in Figure 2. New 2015 ε Nd profiles fall well within the range defined by the profiles collected in the western CB in 2000 (Porcelli et al., 2009).

Spatial and Temporal Variability in 230 Th and 231 Pa Concentrations and Seawater ε Nd
Overall, none of the 230 Th profiles collected after 2000 exceeds the concentration levels of the published, pre-2000s, deep profiles; they are equally high or significantly lower. This observation suggests that the intensity of particulate scavenging in the Amerasian Basin has, in the last two decades, been maintained or increased depending on the area, preventing 230 Th concentrations from building up due to uranium (U) decay.  (Figure 6b), suggesting that the balance between 230 Th removal by particle scavenging and production from U decay was maintained over that time period. By contrast, at the southern CB coastal stations 2000 (dark red circles) and 500 (purple squares) sampled in 2007, 230 Th concentrations

Journal of Geophysical Research: Oceans
are significantly lower compared to the 1995 profile measured at AO-1 in 1995 (yellow circles; Edmonds et al., 1998). A similar change is observed for 231 Pa between station 2000 (dark red circles) and AO-1 (yellow circles; Figure 6c). These observations indicate increasing particle scavenging in the southern  Table 2).

Variability of ε Nd in the CB
Differences between coastal and interior regions of the CB are also observed in ε Nd profiles (Figure 6d). Stations sampled in 2000 (Porcelli et al., 2009) exhibited significant differences in ε Nd , with station 4 (red diamonds), identified as a transition zone station from its 230 Th profile (Figure 6b), having significantly more radiogenic values than station 3 (interior basin). This observation suggests that in 2000, these two locations were in areas of distinct circulation, with little lateral exchange between them. The southern, more radiogenic 2000 profile resembles the one observed at CB2 in 2015, on the eastern CB margin (red diamond and green square in Figure 6d  shallower than 50 m). In a steady state system, the slope of this red line reflects differences in the relative adsorption of Pa and Th. Steeper slopes reflect more intense scavenging of Th (higher k a ) than for Pa. This is discussed further in the text.

Contrasting 230 Th, 231 Pa, and ε Nd Variability in the 2015 CB
The two 2015 deep stations (CB3 and CB4; Figure 6a) exhibit very similar profiles in ε Nd (Figure 6d) and 231 Pa (Figure 6c), but not in 230 Th (Figure 6b). For the latter, the Atlantic layer at CB4 (white circles) exhibits much lower concentrations than at CB3 (black diamonds), resulting in higher 231 Pa/ 230 Th ratios (Figures 8a-8b). These differences reveal ongoing processes for which 230 Th is more sensitive than 231 Pa and ε Nd , most likely an increase in particle fluxes that removes 230 Th faster than the two other geochemical tracers, which are less susceptible to scavenging. CB4 is in closer vicinity to the boundary circulation and exhibits 230 Th concentrations in the Atlantic water similar to those found at the 2015 coastal station CB2 (green squares). However, unlike CB4, CB3 also has lower 231 Pa concentrations and more radiogenic ε Nd . These results suggest that in 2015, the coastal region of the CB is impacted by higher fluxes of particles, which started to affect CB4. It seems to also be the case (3) an area north of 75°N in the central basin subject to low exchanges with the boundary circulation, with more negative ε Nd and a particle flux maintained at relatively low level.

Variability of 230 Th and 231 Pa in the MB
In the Makarov Basin and adjacent ridges area (hereafter MBAR), the highest dissolved 230 Th and 231 Pa concentrations were measured over Alpha Ridge in 1983 (red circles; Bacon et al., 1989) and the northern MB in 1991 (green circles; Scholten et al., 1995; Figure 7). Building up such high concentration requires a residence time exceeding several decades in a region subjected to a very low particle flux region, as would be expected under permanent sea ice cover, to allow ingrowth of the two nuclides from their parent U isotopes. Dissolved 231 Pa/ 230 Th ratios at these two stations are not as high as generally observed in the CB (Figures 8c-8d), which is also consistent with a multidecadal residence time under permanent sea ice. As a water mass enters a region subjected to a lower particle flux, its 230 Th concentration build up faster than 231 Pa concentration because of the differences in production rate and residence time with respect to scavenging between the two radionuclides, resulting in a gradually decreasing dissolved 231 Pa/ 230 Th ratio with time. However, the MB profile comparison highlights (1) Figure 7a). The comparison of the two Alpha Ridge profiles shows significantly lower concentrations of 231 Pa at the 2007 station, even more pronounced for 230 Th, surprisingly tending toward CB concentrations with depth (pink square profile in Figure 7). In the following, we further develop the hypothesis presented above for the CB and hypotheses to explain the MBAR features.

Increased Particle Flux in the Margin Area
In the CB, the low 230 Th and 231 Pa profiles of the two 2007 coastal stations compared to the coastal 1995 profile reveal the impact of an enhanced vertical particle flux. The 2007 stations were located on the margin close to the Mackenzie River mouth, an area more and more seasonally free of ice ( Figure S1): particulate riverine material and biological productivity could both explain an enhanced vertical particle flux at these coastal stations. The deeper 2007 coastal stations (stations 2000 and 2700) exhibit lower dissolved oxygen concentrations than measured in 1995 at station AO-1, supporting an increasing flux of biogenic particles (Figures 9a-9b). Ice rafted particles may also be released during the seasonal ice melt and possibly reinforce the particle flux and scavenging rate in this margin area (Baskaran et al., 2003;Trimble et al., 2004).
In addition to enhanced vertical fluxes of particles in the margin area by increased biological productivity, our 2015 results also suggest lateral fluxes of continental particles. As noted above, the 2015 profiles measured at the two deep stations CB3 and CB4 have similar ε Nd (Figure 6d) and 231 Pa (Figure 6c) concentrations in the Atlantic layer, but lower 230 Th concentration at CB4 (Figure 6b), while at the coastal station 10.1029/2019JC015265

Journal of Geophysical Research: Oceans
(CB2), ε Nd was more radiogenic and both 230 Th and 231 Pa concentrations were lower than at CB3. Comparing the 230 Th profiles to particulate iron (pFe) measured at the same stations ( Figure 10; Li, 2017) reveals that the 230 Th minimum observed in the Atlantic layer at CB4 coincides with a maximum of pFe, both of which are not observed at CB3 (Figures 10b-10c). These observations suggest scavenging by lithogenic particle at CB4, possibly coming from the margin or the Northwind Ridge (this ridge rises tõ 520 m below the surface, 250 km west of CB4; see Figure 3a). Moreover, the Atlantic layer at the coastal station CB2 exhibits much higher pFe and much lower 230 Th concentrations than at the deep stations (Figure 10a), suggesting higher flux of coastal particles that scavenges both 230 Th and 231 Pa. The more radiogenic ε Nd below 500 m at CB2 compared to CB3 and CB4 (Figure 6d) further suggests that this signature is imparted from margin sediments, which is consistent with the ε Nd of the acetic acid leachates of surface sediments measured in the area (Haley & Polyak, 2013). The suggested impact of coastal particles on 230 Th, 231 Pa, and ε Nd data at CB4 cannot only be the result of coastal processes transmitted offshore by advection. Instead, coastal particles must be advected from the margin to CB4 to explain the pFe data. The transport of dissolved and particulate lithogenic signature from nepheloid layers formed by sediment resuspension along the CB margin across the area of boundary circulation could explain the distribution of all these four tracers. The occurrence of nepheloid layers has indeed been reported in the CB margin area (e.g., Ehn et al., 2019;O'Brien et al., 2006O'Brien et al., , 2013. Coastal water and particles could also be

Journal of Geophysical Research: Oceans
transported through eddies generated in the southern CB boundary current (Watanabe, 2011;Zhao et al., 2014), whose number has intensified over the last two decades (Zhao et al., 2016).
The high dissolved 231 Pa/ 230 Th activity ratios in the Atlantic layer of most CB coastal stations collected after 2000 (Figures 8a-8b) are consistent with the preferential scavenging of 230 Th by lithogenic particles (Chase et al., 2002). Such high 231 Pa/ 230 Th activity ratios are not observed in the MBAR area (Figures 8c-8d) where dissolved oxygen concentration profiles are not decreasing with time as in the CB (Figures 9c-9d), suggesting that fluxes of lithogenic and biogenic particles are lower and more constant in time in the MBAR area. This difference is consistent with the evolution of sea ice coverage, more persistent in the MBAR area than in the CB ( Figure S1), limiting biological productivity and ice rafted debris release.

Spatial Characterization of an Area of Lateral Exchanges Between the Boundary Circulation and Northern Central CB
The nonlinear profiles observed in 2000 (station 4), 2007 (station 2700), and 2009 (station L2; Figure 6b) document the presence of an area where coastal low-230 Th and offshore high-230 Th waters influence the 230 Th profiles to different extents at different depths. At these three stations, the relatively high concentrations measured at intermediate depths reflect advection or mixing of water originating from the northern CB. The southern limit of this mixing area in 2007 is fairly well captured by the sharp difference in 230 Th concentrations and potential temperature within the Atlantic layer between the most offshore station 2700 and the more inshore station 2000 located 51 nautical miles south. At station 2700, the Atlantic water has higher 230 Th concentrations (Figure 6b), a higher temperature maximum (Figure 3c; inset), and a temperature profile showing zigzags indicating weak turbulence and doublediffusive intrusions (Woodgate et al., 2007)-consistent with an offshore northern CB origin (McLaughlin et al., 2009). In contrast, the low 230 Th concentration and smooth θ-S profile in the Atlantic layer of station 2000 is consistent with a boundary origin (Figures 5c and 6b). These two profiles are likely located across the front separating the cyclonic boundary current from the central anticyclonic flow (McLaughlin et al., 2009). The 1995 230 Th profile did not exhibit such deviations from linearity and was therefore likely located south of this front.
Interestingly, it seems that the location of this front has moved further north between 2007. Station L1.1, sampled in 2009, is almost at the same location as station 2700 sampled in 2007 (most offshore station; pink vs. green profiles in Figure 11a). It is difficult to explain the sharp decrease observed in 230 Th concentrations within the Atlantic layer between 2007 and 2009 only as a result of increased particle flux and scavenging, considering the residence time of Th at these depths. Instead, it most likely reflects a

10.1029/2019JC015265
Journal of Geophysical Research: Oceans greater relative proportion of "margin" versus "offshore" water at this location in 2009. The smoother θ-S within the Atlantic layer and more radiogenic, 230 Th-depleted signal found at 1,000 m at the 2015 station CB4 (white circles, Figure 6), compared to the nearby 2000 station 3 (orange circles; Figure 6) sampled 15 years before may also reflect the northward displacement of this front and increase lateral exchanges between the margin area and the central CB during the last two decades (orange vs. brown profiles in Figures 11a and 6d). Figure 11b summarizes, through the characteristics of the 230 Th vertical profiles, the spatial distribution and temporal variability of the impact of coastal processes and lateral exchanges.
The lateral exchange variability suggested in the CB may be related to the propagation in the Arctic of the awAW: associated with a larger volume of inflowing Atlantic water (Karcher et al., 2003;Schauer et al., 2004), this awAW propagated as a stronger intermediate flow, which presumably reinforced the two circulation schemes of the CB (anticyclonic circulation in the deep basin; Coachman & Barnes, 1963;Newton & Coachman, 1974; boundary cyclonic circulation) and enhanced the lateral mixing between the cyclonic boundary circulation and the anticyclonic central circulation. This is supported by the coexistence of the temporal increase of vertical variability within the 230 Th profiles with the temporal increase of temperature of the upper Atlantic layer (Figure 11a). The eddy activity and its intensification over the last decades (Timmermans et al., 2008;Watanabe, 2011;Zhao et al., 2016) could also be partly responsible for the lateral exchange variability suggested by our profiles.  (Figure 12). This suggests the relative isolation of the water column in AR area from the MB and CB circulation. In addition, the high 230 Th d and 231 Pa d concentrations measured in 1983 in the northern AR (station CESAR) reveal the long isolation of the water column in this area under low flux of particles, allowing ingrowth of the two radionuclides (see Figure 2c; Bacon et al., 1989). In 1996, Smethie et al. (2000 consistently found low CFC-11 concentrations and 3 H-3 He ages of 25 years on average below 300 m in this area. From the comparison of the hydrological profiles at these two AR stations to those of the whole station set reported here (Figures 3-5 and 9), AR waters appear to be dominantly derived from CB waters and, more particularly, from colder, saltier, and denser waters predating the awAW propagation in the CB. The largest hydrological differences observed between the AR stations collected in 1983 and 2007 are actually not in the AW core (~500 m) but in the lower halocline and lower AW, where the 2007 station is warmer (Figure 12b). These geochemical and hydrological features of the two AR stations are consistent with the lower AW and uPDW in 2007 being relatively recently renewed by CB waters. The 230 Th d and 231 Pa d profiles are consistent with this hypothesis, as well as the dissolved oxygen (Figures 7 and 9d Timmermans et al., 2003) in 2015 in the MB (stations 96, 101, and 134; Figures 7 and 12c). Indeed, the deep and bottom layers of the CB are characterized by significantly lower concentrations of 230 Th than the MB; the uniformity in the 231 Pa concentrations of the MB and northern CB at this depth also support such exchanges. In addition, the MB 231 Pa/ 230 Th activity ratios show a consistent deviation around this depth toward slightly higher, CB-like, values (Figure 8c). The fact that this 230 Th minimum is clearly observed in the eastern MB (station 96) further suggests that this CB signature spreads through the entire Basin (i.e., east of 180°). Hydrological features such as the θ-S convexity and O 2 minimum observed within the DTL at the MB stations also support exchanges of CB deep waters toward the MB (Figures 4, 5h, and 9d). Although the direction of exchanges we suggest here conflicts with former hydrological study conclusions that invoked a dominant flux of DTL water above the Mendeleev/Alpha Ridge from the MB toward the CB (e.g., Carmack et al., 2012;Rudels, 2009), Swift et al. (1997 also invoked a flow from the CB toward the MB above the Mendeleev Ridge to explain the local silicate maximum observed in the MB around 2,400 m depth.

Lateral Exchanges in Other
Intrusions of CB deep waters into the MB could also explain the warmer and saltier water observed in the bottom waters at the MB and Mendeleev Ridge stations after 2000 (insets in the θ-S plots in Figures 12b-12c). Nonetheless, considering the gradual decrease of tracer-especially 231 Pa d -concentrations with depth below 3,000 m, coupled to a gradual increase of dissolved O 2 (Figures 9b and 12c), we agree with Middag et al. (2009) and Roeske et al. (2012) that MB bottom water is also made of some colder and fresher Amundsen Basin Deep Water (ABDW) that likely came over the Lomonosov Ridge (between 2,000 and 2,500 m) and sank due to its higher density than the local bottom waters (Timmermans et al., 2005).

Lateral Exchanges Between the Eurasian Basin and the Central Makarov Basin
As previously mentioned, the data reported in this work suggest that fluxes of lithogenic and biogenic particles have been weaker and more constant in time in the MBAR area than in the CB (see section 5.2). Yet significant differences are observed between the 230 Th d and 231 Pa d profiles in the MB before 2000 (Station 176; Scholten et al., 1995) and those at the stations visited after 2000 ( Figure 12c): post-2000s profiles exhibit lower radionuclide concentrations, especially pronounced in 231 Pa d , compared to the 1991 profile. The high 230 Th d and 231 Pa d concentrations measured in 1991 in the northern MB reflect, as for the 1983 station collected over the AR, a long isolation of the area from the dynamic boundary circulation of the MB and a low flux of particles (Scholten et al., 1995). The difference of 231 Pa d concentration between this 1991 station and those visited after 2000 increases with depth. It is difficult to explain the pre-versus post-2000s radionuclide differences observed in the MB as a stronger flux of particles, as a change in particle flux strong enough to explain the 231 Pa changes in depth should have also led to a much larger 230 Th depletion than those observed ( Figure 12c). Not to mention that, despite the evolution of sea ice coverage since the 1990s, most of the MB has remained permanently covered by sea ice over these years ( Figure S1), limiting an enhancement of productivity.
Therefore, our geochemical results suggest that waters with lower 230 Th concentrations and lower 231 Pa/ 230 Th activity ratios (due to lower concentrations of 231 Pa relatively to 230 Th) propagated into the MB after 1991. CB waters are characterized by too high 231 Pa/ 230 Th activity ratios and too low oxygen concentrations to represent a valid candidate. In contrast, consistent lower 230 Th concentrations and lower 231 Pa/ 230 Th activity ratios have been documented in the Eurasian waters (Edmonds et al., 2004), as well as similar oxygen concentrations (Schauer et al., 2002). The presence of the awAW at the stations visited in the MB after 2000 supports the occurrence of exchanges between this previously isolated area and the topographically steered intermediate circulation (Figure 12c).

Conclusions
This work presents a compilation of published and new depth profiles of 230 Th, 231 Pa, and ε Nd collected between 1983 and 2015 in the Amerasian Basin of the Arctic Ocean, in particular in the Canada and Makarov basins and over the Mendeleev and Alpha Ridges. The distribution of these geochemical tracers allow an assessment of the spatial and temporal variability in particle flux and water mass circulation and 10.1029/2019JC015265 Journal of Geophysical Research: Oceans mixing in these different areas. A representation of intermediate-to-bottom layer circulation, including pathways likely dominant in the past but suggested to be more minor recently (section 5.4) as well as the circulation from the main conclusions (summarized below) is outlined in Figure 13.
A temporal decrease in concentrations of the particle-reactive tracers 230 Th and 231 Pa is observed in the whole Amerasian Basin and results from particle scavenging, and mixing and circulation variability. Particularly in the Canada Basin, the intensification of lithogenic and biologic particle fluxes likely results from margin sediment transport and enhanced biological production, respectively, in relation to sea ice retreat.
Imprints of increased lateral exchange are also reported in several areas of the Amerasian Basin. In the Canada Basin, they are represented by the coexistence of high 230 Th Atlantic water of northern origin overlying low 230 Th waters that have their origin in the boundary; the spatial extent of this area of lateral exchange seems to vary in time. The geochemical and hydrological characteristics at the southern flank of the Alpha Ridge in 2007 reflect the occurrence of lateral exchange in the Atlantic layer and uPDW with Canada Basin waters in the early 2000s, in contrast with the isolated character of this area reported in the 1980-1990s. Similarly, the geochemical and hydrological characteristics reported in the MB suggest that lateral exchange has connected the previously isolated northern MB with the main circulation (the pre-2000s Makarov Basin circulation was likely restricted to the red arrow shown in Figure 13a; blue arrows show the connection of the offshore Makarov Basin with the boundary circulation suggested from the post-2000s profiles). The results reported in this study also suggest increased intrusion of Canada Basin deep waters over the Mendeleev-Alpha Ridge sill into the Makarov Basin, impacting the Deep Transitional Layer (2,000-2,500 m) waters of the Makarov Basin (while a dominant overflow from the Makarov Basin toward the

Journal of Geophysical Research: Oceans
Canada Basin was previously suggested, as represented by the red arrow in Figure 13b). Canada Basin Deep Water intrusion also seems to mix with deeper waters of the Makarov Basin and decrease the Amundsen Basin Deep Water (ABDW) signature of the Makarov Basin Bottom Water (ABDW overflow represented by the red arrow in Figure 13c).
These conclusions are based on sparse samples, in space and time. Additional sampling is needed to refine the hypotheses proposed in this study. There is a need to better document the coastal Makarov Basin, from the Lomonosov Ridge to the Mendeleev Ridge, and the northern Canada Basin, from the Chukchi Plateau to the northeast Canadian Arctic Archipelago (CAA). It would be of interest to sample the Alpha Ridge again over the southern and northern flanks to determine whether this area is isolated, or if changes in the Arctic circulation have led to exchange of waters between this area and the Canada Basin. Finally, with the circulation changes suggested in the Makarov Basin, it would also be interesting to monitor the water circulation in the vicinity of the western Lomonosov Ridge, by sampling the Greenland side of the Makarov Basin, Alpha Ridge and Amundsen Basin, to better document the composition of waters flowing through Nares Strait (between the CAA and Greenland) and those returning to Fram Strait. The circulation and particle distribution of the Amerasian Basin are sufficiently heterogeneous to make these geochemical tracers a useful tool to study the Arctic Ocean state and evolution, though the utility of these tracers may be challenged if mixing and particle concentration dramatically increase and eventually erases the current gradients.