Dissolved Pb and Pb isotopes in the North Atlantic from the GEOVIDE transect (GEOTRACES GA-01) and their decadal evolution

. During the 2014 GEOVIDE transect, seawater samples were collected for dissolved Pb and Pb isotope analysis. 15 These samples provide a high resolution “snapshot” of the source regions for the present Pb distribution in the North Atlantic Ocean. Some of these stations were previously occupied for Pb from as early as 1981, and we compare the 2014 data with these older data some of which are reported here for the first time. Lead concentrations were highest in subsurface Mediterranean Water (MW) near the coast of Portugal, which agrees well with other recent observations by the U.S. GEOTRACES program (Noble et al., 2015). The recently formed Labrador Sea Water 20 (LSW) between Greenland and Nova Scotia is much lower in Pb concentration than the older LSW found in the Western European Basin due to decreases in Pb emissions into the atmosphere during the past 20 years. Comparison of North Atlantic data from 1989 – 2014 shows decreasing Pb concentrations consistent with decreased anthropogenic inputs, active scavenging, and advection/convection. The nearly-homogenous isotopic composition of northern North Atlantic seawater implies that the relative proportions of U.S. and European Pb sources to the ocean 25 have been relatively uniform during the past two decades. Using our measurements in conjunction with emissions inventories, we support the findings of previous atmospheric analyses that up to 50% of the Pb deposited to the ocean in 2014 was natural, although it remains unclear if that natural dust is from the mid- or high-latitudes. Atlantic, another 2010 - 2011 study found that 50 – 70% of Pb in the surface ocean was anthropogenic in origin (Bridgestock et al., 2016), with the remaining fraction from natural North African dust. This study evaluates current sources and relative quantities of Pb in the northern North Atlantic Ocean. We compare 25 these findings with older seawater Pb data (some published for the first time here). Our study is strongly enhanced by the partnership of the environmental trace metal GEOTRACES program with the OVIDE program’s long-term studies of physical oceanographic parameters in the Northeast Atlantic (Garcia-Ibanez et al., 2015). Pb concentration analysis for 1989 (Atlantis II 123) was achieved by 204 Pb isotope dilution with Mg(OH) 2 coprecipitation followed by VG PQ2+ quadrupole ICPMS (Wu & Boyle, 1997) (analysed in and 1999 (Endeavor 328) stations 4, 7, 9, 10 and 11 (analysed between 1999-2003). Endeavor 328 stations 2, 3, 8, and 10 were determined using NTA-extraction ID ICPMS (Lee et al., 2011) (determined in 2010). Long-term quality control seawater samples were included in each run, and overlapped with new QC samples when the previous 5 depleted. Endeavor 328 station 10 was determined twice by two analysts 8 years apart (in 2002 by Mg(OH) 2 coprecipitation ID-ICPMS, and in 2010 by NTA extraction ID-ICPMS). A regression of the 2010 vs 2002 We provides our demonstrates during minutes concentrated ultrapure HNO 3 of 400µL of ultrapure with an appropriate amount of Tl for mass fractionation correction. IsoProbe multicollector ICPMS Faraday cups were used to collect on 202 Hg, 203 Tl, 205 Tl, 206 Pb, 207 Pb, and 208 Pb. An Isotopx Daly detector with a WARP filter was used to collect on 204 Pb+ 204 Hg. the deadtime of the Daly detector varied from day to day, we calibrated deadtime each day a 206 Pb/ 204 Pb at a high 204 count rate. Pb/ Pb (and 204 count rate 209 Bi half-mass sample 30 measured NBS981 ratios determinations of Pb isotope for samples of the Pb shown for samples (~1V), 206 Pb/ Pb and Pb/ 207 Pb samples be worse than but generally better than 1000ppm in data uncertainty in Pb/ Pb for samples in the mid-to-upper range of sample concentrations reproducible at Pb near-bottom rosette position than 1, Pb over the samples immediately above. We believe that this evidence 30 points to GoFLO bottle-induced contamination that was being slowly washed out during the cruise, but never completely. A similar pattern was observed for the samples taken from rosette positions 5, 20 and 21 when compared to the depth-interpolated [Pb] from the samples immediately above and below. We do not believe that these samples should be trusted as reflecting true ocean [Pb], so all of the samples from these GOFlos are excluded in our discussion of this work, although they are included and flagged as unreliable within the data repositories.

In the western North Atlantic, repeat sampling of time series locations have documented the reduction in oceanic [Pb] and changes in sources with time. At BATS (Bermuda Atlantic Time Series) in the 1970s and 1980s, 15 concentrations were 80 -160 pmol kg -1 near the surface but 25 pmol kg -1 at depth (Boyle et al., 2014 and references therein). As Pb emissions were reduced and surface waters subducted, the elevated [Pb] could be seen as a plume in subsurface waters at increasingly deeper depths over time. At the latest occupation of BATS (2011), surface water concentrations were less than 20 pmol kg -1 (Noble et al. 2015). Despite a dramatic reduction in [Pb], it is still believed that a large fraction of that Pb is a result of (coal) combustion and industrial processes based on positive 20 matrix factorisation analysis of aerosols (Shelley et al., 2017;Noble et al 2015). In the tropical Atlantic, another 2010 -2011 study found that 50 -70% of Pb in the surface ocean was anthropogenic in origin (Bridgestock et al., 2016), with the remaining fraction from natural North African dust.
This study evaluates current sources and relative quantities of Pb in the northern North Atlantic Ocean. We compare 25 these findings with older seawater Pb data (some published for the first time here). Our study is strongly enhanced Intelligent Rosette), equipped with new, cleaned 12L GO-FLO bottles (Cutter and Bruland, 2012). The rosette was deployed on a 6mm Kevlar cable with a dedicated custom designed clean winch. Immediately after recovery, GO-FLO bottles were individually covered at each end with plastic bags to minimize contamination. They were then transferred into a clean container (class-100) for sampling. For Stations 1,11,15,17,19,21,25,26,29,32 samples were filtered with 0.2 µm capsule filters (SARTOBRAN® 300,Sartorius). For all other stations (13, 34, 36, 38, 40, 5 42, 44, 49, 60, 64, 68, 69, 71, 77) seawater was filtered directly through paired filters (Pall Gelman Supor 0.45µm polystersulfone, and Millipore mixed ester cellulose MF 5 µm) mounted in Swinnex polypropylene filter holders, following the Planquette and Sherrell (2012) method. All samples were acidified back in the MIT laboratory with 2mL trace metal clean 6M HCl per liter of seawater (final pH ~2). 10 Previously unpublished Pb and Pb isotope data from cruises from 1989 (Atlantis II cruise 123) and 1999 (cruise Endeavor EN328) are included here for evaluation of the decadal evolution of Pb in the eastern North Atlantic. We supplement our 1989 data with two published JGOFS stations (Martin et al., 1993). Our 1989 samples were collected using "vane bulb" samplers (Boyle et al., 1986) and the 1999 samples were collected using the MITESS mooring sampler (Bell et al., 2002). Samples were stored in acid-cleaned 250ml LPE bottles. 15

Pb Concentrations
GEOVIDE samples were analysed at least 1 month after acidification during more than 36 analytical sessions using the isotope-dilution ICP-MS method described in Lee et al. 2011, which includes pre-concentration on nitrilotriacetate (NTA) resin and analysis on a quadrupole ICP-MS (Fisons PQ2+). Method details including all 20 cleaning protocols are available in the metadata file, along with the data, in the BCO-DMO repository (see 2.4).
Briefly, triplicate subsamples (1.3mL) were spiked with a known 204 Pb spike and the pH was raised to 5.3 using a trace metal clean ammonium acetate buffer, prepared at a pH of between 7.95 and 7.98. ~2400 beads of cleaned NTA Superflow resin (Qiagen Inc., Valencia, CA) were added to the mixture and equilibrated. After equilibration, the resin was rinsed with distilled water and then Pb was eluted with a 0.1M solution of trace metal clean HNO 3 25 before analysis by ICP-MS.
On each day of sample analysis, procedural blanks were determined for 12 replicates of in-house reference seawater with negligible [Pb]. The blanks analysed concurrently with these samples ranged from 2.2 -9.9 pmol kg -1 , averaging 4.6 ± 1.7 pmol kg -1 . Within a day, procedure blanks were very reproducible with an average standard deviation of 0.7 pmol kg -1 , resulting in detection limits (3x the low-level standard deviation) of 2.1pmol kg -1 . 30 Replicate analyses of three different large-volume seawater samples (one with ~11 pmol kg -1 , another with ~24 pmol kg -1 , and a third with ~38 pmol kg -1 ) indicated that the precision of the analysis is 4% or 1.6 pmol kg -1 , whichever is larger. Triplicate analyses of an international reference standard SAFe D2 were 27.2 ± 1.7 pmol kg -1 .
Pb concentration analysis for 1989 samples (Atlantis II 123) was achieved by 204 Pb isotope dilution with Mg(OH) 2 coprecipitation followed by VG PQ2+ quadrupole ICPMS (Wu & Boyle, 1997) (analysed in 1996(Endeavor 328) stations 4, 5, 7, 9, 10 and 11 (analysed between 1999. Endeavor 328 stations 2, 3, 8, and 10 were determined using NTA-extraction ID ICPMS (Lee et al., 2011(Lee et al., ) (determined in 2010. Long-term quality control seawater samples were included in each run, and overlapped with new QC samples when the previous 5 samples were depleted. Endeavor 328 station 10 was determined twice by two analysts 8 years apart (in 2002 by Mg(OH) 2 coprecipitation ID-ICPMS, and in 2010 by NTA extraction ID-ICPMS). A regression of the 2010 vs 2002 data forced through the origin had a slope of 0.945. We suggest that this small offset provides a reasonable estimate of our inter-decadal analytical reproducibility. It also demonstrates that Pb is not continuously leached from wellcleaned LPE bottles during decadal-scale storage. 10

Stable Pb Isotopes
GEOVIDE samples were analysed during 11 mass spectrometry sessions by the method of Reuer et al. (2003) as modified by Boyle et al. (2012). In brief, ~500mL of seawater was pre-concentrated using a low-blank double magnesium hydroxide co-precipitation, induced by minimal addition of high-purity ammonia solution and mixing 15 (typically 8µL ammonia per 1mL seawater sample). The precipitate was dissolved in a minimal amount of highpurity 6M HCl before undergoing another ammonia addition and second Mg(OH) 2 coprecipitation. The final precipitate was dissolved in ~1mL of high purity 1.1M HBr the day of purification by anion exchange chromatography (Eichrom AG1x8). Samples were dried and stored in PTFE vials until isotope ratio analysis on a GV/Micromass IsoProbe multicollector ICPMS using an APEX/SPIRO desolvator. Just before analysis, samples 20 were dissolved for several minutes in 10µl concentrated ultrapure HNO 3 followed by addition of 400µL of ultrapure water and spiked with an appropriate amount of Tl for mass fractionation correction. IsoProbe multicollector ICPMS Faraday cups were used to collect on 202 Hg,203 Tl,205 Tl,206 Pb,207 Pb,and 208 Pb. An Isotopx Daly detector with a WARP filter was used to collect on 204 Pb+ 204 Hg. Because the deadtime of the Daly detector varied from day to day, we calibrated deadtime on each day by running a standard with known 206 Pb/ 204 Pb at a high 204 count rate. The 25 counter efficiency drifts during the course of a day, so we established that drift by running a standard with known 206 Pb/ 204 Pb (and a 204 count rate comparable to the samples) every five samples. Tailing from one Faraday cup to the next was corrected by the 209 Bi half-mass method as described by Thirlwall (2000).
On each analytical date, we calibrated the instrument by running NBS981 and normalized measured sample isotope 30 ratios to our measured raw NBS981 isotope ratios to those established by Baker et al. (2004). Using this method for 22 determinations of an in-house Pb isotope standard solution shows that for samples near the upper range of the Pb signals shown for samples (~1V), 206 Pb/ 207 Pb and 208 Pb/ 207 Pb were reproduced to ~200ppm. Low-level samples will be worse than that, but generally better than 1000ppm in this data set. Because of the drift uncertainty in the Daly detector, 206 Pb/ 204 Pb for samples in the mid-to-upper range of sample concentrations will be reproducible at best to 35 ~500ppm.
We have intercalibrated Pb isotope analyses with two labs as reported in Boyle et al. (2012). The outcome of that intercalibration suggests that the accuracy of our measurements approaches the internal analytical reproducibility we note above.
Pb isotope precision for the complete analytical procedure can be assessed by duplicate measurements of samples. In 5 most cases, the replicated samples were chosen because they fell off of the trend of adjacent samples. That could be due either to contamination of the subsample used for the analysis, or to the contamination of the sample in its primary sample bottle. As shown in Figure S2, the replicate analysis usually agreed within better than 1000ppm for 206 Pb/ 207 Pb and 208 Pb/ 207 Pb, and 5000ppm for 206 Pb/ 204 Pb. We suggest that these provide a reasonable upper limit for the replicability of our isotope measurements. 10 Pb isotope data from the 1999 samples were obtained by IsoProbe Multicollector ICPMS after Mg(OH) 2 preconcentration and anion exchange purification as described by Reuer et al. (2003). As for the GEOVIDE samples, the mass spectrometer was calibrated using NBS981.

Data Management
All [Pb] and isotope data related to the GEOVIDE data set in this manuscript have been submitted to BCO-DMO and will be available at 〈http://www.bco-dmo.org/dataset/651880/data and http://www.bcodmo.org/dataset/652127/data〉 and from the 2017 BODC International GEOTRACES Intermediate Data Product v2 (The GEOTRACES Group, 2015). All other data is available in table 1. 20

Outliers
In this data set, we did not encounter any samples that did not yield acceptably reproducible results upon repeated analysis, so we believe that the data truly represents the concentration of Pb in the sample collection bottle.
However, there were a few samples with elevated Pb based on adjacent samples and for which no obvious 25 hydrographic argument could be made for the anomaly. We observed that the samples taken from GOFlo in rosette position 1 (usually the near-bottom sample) were always higher in [Pb] than the samples taken immediately above that, and that the excess decreased as the cruise proceeded ( Figure S1). The Pb isotope ratio of these samples were higher than the comparison bottles as well. At two stations where our near-bottom sample was taken from rosette position 2 rather than 1, there was no Pb excess over the samples immediately above. We believe that this evidence 30 points to GoFLO bottle-induced contamination that was being slowly washed out during the cruise, but never completely. A similar pattern was observed for the samples taken from rosette positions 5, 20 and 21 when compared to the depth-interpolated [Pb] from the samples immediately above and below. We do not believe that these samples should be trusted as reflecting true ocean [Pb], so all of the samples from these GOFlos are excluded in our discussion of this work, although they are included and flagged as unreliable within the data repositories. 35 Biogeosciences Discuss., https://doi.org /10.5194/bg-2018-29 Manuscript under review for journal Biogeosciences Discussion started: 2 February 2018 c Author(s) 2018. CC BY 4.0 License.
In addition, we observed high [Pb] in most of the samples from Station 1 and very scattered Pb isotope ratios. The majority of these concentrations were far in excess of those values observed at nearby Station 11, and also the nearby USGT10-01 (Noble et al., 2015). Discussion among other cruise participants revealed similarly anomalous data for other trace metals (e.g., Hg species; personal communication with L.-E. Heimburger). After discussion at 5 the 2016 GEOVIDE post-cruise workshop, we came to the conclusion that this is evidence of GoFlo bottles not having sufficient time to "clean up" prior to use, and that most or all bottles from Station 1 were contaminated.
Station 1 data is not discussed in this work, but as with the suspicious GOFlos throughout the cruise, the Station 1 data are included and flagged as unreliable in the data repositories.

Near-surface Ocean 10
Near-surface waters (11 -20 m) displayed a moderate range in [Pb] across the transect ( Figure 2). The highest concentration was located near the Portugal coast (30 pmol kg -1 ). Lead concentrations decreased three-fold with distance from the coast, down to 11.5 pmol kg -1 , in the core of the far arm of the North Atlantic Current. An excellent pictorial representation of the relevant water masses discussed here can be found in Garcia-Ibanez et al. The pattern of decreasing [Pb] over the Iberian Abyssal Plain (Stations 11 -19) correlates strongly with increasing distance from shore (Pearson's correlation, r = -0.989, p < 0.001). This finding agrees well with atmospheric 20 deposition models that show higher dust inputs closer to the African continent (Schepanski et al., 2009). Stations located north of 55° in the meandering NAC have higher concentrations than those in the Western European Basin.
Although dust deposition to the North Atlantic Ocean is typically associated with North African dust from the Saharan Desert, Prospero et al. (2012) and Bullard et al. (2016) found that high latitude dust emissions, specifically volcanic-based soils from Iceland, could be substantial enough to impact oceanic Fe cycling; therefore we suggest 25 that the elevated Pb in the near-surface waters of the Icelandic Basin and Irminger Sea may possibly be dust-derived.
In the GEOVIDE shipboard aerosol data (Shelley et al., 2017), Pb concentrations were high in Iceland Basin but low in the Irminger Sea. However, as Pb has a residence time of ~1 year in this region, seasonal changes in the flux could account for this discrepancy. As the North Atlantic Current becomes the Irminger Current near Greenland and joins with the East Greenland Current, they wrap around the southern tip of Greenland and flow toward the Arctic 30 Circle. This entrains Pb into the northeast part of the Labrador Sea, whereas the remainder of the Labrador Sea is influenced by the Labrador Current, returning from the Arctic, which has low [Pb].
Despite the variations in [Pb] across the Atlantic Ocean, Pb isotope ratios were relatively homogenous throughout the section, and largely decoupled from the [Pb] patterns (Figures 3, 4). 206  showed similar minimal variability. No trend in isotope ratios was observed in the Iberian Abyssal Plain extending away from the coast. The low variability of isotope ratios indicates that the majority of Pb in the Northern Atlantic Ocean is well mixed in the atmosphere prior to deposition. The relatively low [Pb] and similar isotope ratios contrast sharply with surface water measurements from the previous century ( Figures 5, 6). During the 1970s -early 1990s, the predominant source of Pb to the North Atlantic was U.S. leaded gasoline (Weiss et al, 2003;Martin et al., 1993;5 Veron et al., 1999), which was reflected in the high 206 Pb/ 207 Pb isotope ratios (~1.20).
The mixed layer [Pb] nearest the Iberian Peninsula (30 pmol kg -1 ) are lower than those measured by the 2010 US GEOTRACES expedition (42 pmol kg -1 ), which we attribute to the much closer proximity of the US GEOTRACES station to the coastline (50 km) than GEOVIDE station 11 (280 km). As mentioned previously, [Pb] at GEOVIDE 10 stations 11 -19 have a strong inverse correlation to distance from shore, and adding USGT10-01 (GA-03) maintains this high correlation (Pearson correlation, r = -0.990, p < 0.001). Isotopically, the USGT10-01 near-surface waters are similar to GEOVIDE station 11, indicating similar Pb sources in recent years.

Iberian Abyssal Plain (S11 -S19) and Western European basin (S21 -S29)
Overall, [Pb] Noble et al. (2015) and highlights the high [Pb] previously found in the Mediterranean Sea (Moos and Boyle, in preparation). In the lower portion of the plume, the LSW in the Iberian Abyssal Plain and Western European basin is among the oldest water sampled during this expedition. According to CFC-11 data, LSW in this region has a combined age (subduction plus admixed relic age) of ~25 years (Fine, 2011). 25 That age and the elevated [Pb] observed are consistent with the atmospheric Pb emissions by North America and Europe in the 1980s. The isotope ratios further support this finding, as the ocean interior has similar isotope ratios throughout ( 206 Pb/ 207 Pb = 1.1832 +/-0.0025, 1σ; 208 Pb/ 206 Pb = 2.4525 +/-0.0024, 1σ), but are distinguishably more like US aerosols from the early 1990s (Bollhofer & Rosman, 2001) at the core of the Pb maximum (Station 13, 206 Pb/ 207 Pb = 1.1894; 208 Pb/ 206 Pb = 2.4544; Figure 5). 30 The offshore profiles (Stations 13 -29) showed consistent decreases in [Pb] in the MW and LSW from 1989 (JGOFS S19) and 1999 (Endeavor 328 S15, 17, 21) to 2014 (Martin et al., 1993;this work). In the 10 -15 years between sampling events the Pb maxima advected into the ocean interior as the more shallow waters were ventilated with lower-Pb surface waters, a trend also seen in the western North Atlantic near Bermuda .  Figure 6). This makes sense because the estimated age of NEADW is several hundred years (Matsumoto, 2007).
Below 1000m, the [Pb] at Stations 11 and 13 were very similar to the 2010 [Pb] measured on GA-03 (USGT10-01; Figure 5), but the isotope ratios are dissimilar ( Figure 6). Conversely, the upper 1000m of the water column had different [Pb] but similar isotope ratios. In the upper ocean, this discrepancy can be related to the distance of the 10 stations from shore, as calculated in section 3.2, with greater Pb inputs and therefore greater concentrations at stations closer to shore. In the deep ocean, the contrast in isotope ratios between the more coastal GA03 station and offshore GEOVIDE station, only 4 years apart, imply that the two cruises were sampling in different water masses, either because they sampled ~250 km apart or due to a boundary shift that occurred in the 4 years between sampling events.  190 -1.20) and that ISOW reflected atmospheric emissions from Europe at that time. The differences in Pb isotopes (and 25 two to three-fold reduction in concentrations) between sampling campaigns highlights the young age of ISOW, which reflected large source changes over a 21 year time period (Figures 5, 6).
In addition, we note that the present-day Norwegian Sea waters must have low [Pb], and that their Pb isotope ratios reflect a greater contribution from European sources than North American sources. ISOW is formed as a mixture of 30  Veron et al. (1999Veron et al. ( ) observed in 1993, and are indicative of atmospheric Pb from a more European provenance than North American one (Figure 7). pmol kg -1 ) and a higher 206 Pb/ 207 Pb ratio (1.1854) than ISOW, consistent with the 1993 data of Veron et al. (1999;206 Pb/ 207 Pb = 1.179 -1.182). DSOW is a mix of the Nordic Sea waters overflowing the Greenland-Iceland sill and mixing with LSW; DSOW is also reported to have inputs from dense Greenland shelf water and cascading Polar 15 Intermediate Water (Garcia-Ibanez et al., 2015; this study). The resulting DSOW isotope composition is very similar to LSW, which could indicate shelf water has very little Pb and so its signal is dominated by the LSW signal, although we cannot rule out the possibility that the shelf water entrained Pb with a similar isotope composition to LSW. 20 The Irminger Sea was previously sampled for Pb during the 1993 IOC-2 expedition (Veron et al., 1999)

Labrador Sea (S64 -77)
In the Labrador Sea, the Pb maximum coincides with LSW (0 -2500m) and is very broad. Similar concentrations (~25 pmol kg -1 ) are found from 100m to nearly 2000m. At depths greater than 2000m, the [Pb] decreases to ~8 pmol kg -1 and water mass analysis indicates this is primarily ISOW. Throughout the entire Labrador Sea water column Pb

Sources of Pb in 1999 and 2014
Overall, Pb isotope ratios throughout the GEOVIDE expedition were nearly homogenous  (Veron et al, 1999). The near uniformity of Pb isotope ratios in 2014 throughout the water column shows that the relative proportions of isotopically distinct sources of Pb have been similar for the ~20 years preceding sampling.

20
Atmospheric deposition is the main source of Pb to the ocean, with surface waters reflecting the most recent inputs.
Trace metal enrichment factors of dry aerosols and wet deposition were collected during the GEOVIDE cruise (Shelley et al., 2017). Results for Pb enrichment were moderate (>10), suggesting much of the atmospheric Pb was derived from anthropogenic sources. Using positive matrix factorization of the aerosol concentration data, Shelley et al. estimated that ~60% of the Pb was from a mineral dust source and only 40% was of anthropogenic origin. This 25 finding parallels the 2010 study of Pb in the tropical North Atlantic by Bridgestock et al (2016). That group determined 30 -50% of the surface water Pb was of natural mineral dust from the North African dust plume. A triple isotope plot of surface waters from this cruise (Figure 7a) shows visually good agreement with Shelley et al.'s estimate that ~half of the Pb in the surface waters is of natural origin, assuming that the Icelandic dust has a similar isotopic composition as pre-anthropogenic North African dust. 30 The mixed sources of Pb throughout the ocean interior are similarly clear. In Figure 7b Pb is strongly decreasing in the upper ocean during this period, a fact that can be attributed to the phasing out of tetraethyl Pb gasoline in North America and Europe. All three periods show a Pb maximum in the deep thermocline, and this maximum deepens from decade to decade, as it has also done in the western North Atlantic water column near Bermuda (Boyle et al., 2014;Noble et al., 2015). As Noble et al. demonstrated for the 2010/2011 GA-03 trans-North Atlantic section, this maximum is located in waters with SF 6 ventilation dates from the 1970's, when leaded 25 gasoline Pb utilization was at its maximum. A similar result is seen for the 1989 data based on 3 He-3 H dating (Jenkins, 1987). Hence the location of the maximum is dominantly a reflection of Pb emissions at the ventilation age of the water rather than an association with a particular water mass. When considered in this light -as a snapshot of an evolving three-dimensional transient tracer experiment -some of the features in these sections require an interpretation that differs substantially from that usually placed on quasi-steady-state tracers such as salinity, 30 oxygen, and nutrients. For example the [Pb] maximum seen at ~25°N is not the source of a northward-spreading plume, it is the southern extent of high- [Pb] waters that were subducted into the thermocline in the 1970's and advected southwesterly by the dynamics of the ventilated thermocline (Luyten et al., 1983). In addition to the general ventilation of the North Atlantic water column, some [Pb] features are due to specific hydrographic features.
The 1999 [Pb] maximum near 1000m was enhanced by a strong "Meddy" (a coherent mesoscale feature created by 35 pulses of dense salty water from out of the Mediterranean Sea (Armi et al. 1989), as demonstrated by the salinity Biogeosciences Discuss., https://doi.org /10.5194/bg-2018-29 Manuscript under review for journal Biogeosciences Discussion started: 2 February 2018 c Author(s) 2018. CC BY 4.0 License. data from that profile (Figure 11). It is also evident that the ~1800m Labrador Sea water has had consistently higher Pb than the more dense Greenland-Scotland overflow water.
It is likely that the evident decline in the Pb inventory of the eastern North Atlantic is decreasing not only because of advective/diffusive spreading of the water out of the basin, but also because of scavenging. Radiochemical studies 5 (Bacon et al., 1976;) have shown that deep water column 210 Pb activities are lower than 226 Ra activities signifying removal of 210 Pb from the deep water column. Some of this scavenging is due to sinking particles but in near bottom waters, "boundary scavenging" accounts for a higher fraction (Bacon, 1988).
The evolution of the Pb isotope data between 1999 and 2010-2014 is striking in that the deepest waters in the 10 tropical Eastern Atlantic are significantly changed between these periods. Near the surface, recent changes are mainly due to a greater reduction of the relative North American high 206 Pb/ 207 Pb sources relative to the European low 206 Pb/ 207 Pb sources. But in the deepwater, this change probably represents the "conveyor belt" motion of deep high 206 Pb/ 207 Pb introduced from the surface in the early 1900's being replaced by lower 206 Pb/ 207 Pb from the 1920's and later (as seen in historical Pb isotope ratios in Bermuda corals, Kelly et al., 2009). 15

Conclusions
In the past 30 years, massive reductions in Pb emissions to the environment have been evidenced by sampling campaigns in the North Atlantic Ocean. Evolution of [Pb] and Pb isotope ratios will continue as human-derived 20 emissions continually decline, Pb is naturally scavenged from the water column, and the oceanic "conveyor belt" continue to mix deep waters. Like Bridgestock et al. (2016) found in the tropical Atlantic, we see evidence of a natural Pb source to the northern North Atlantic which was previously obscured in the 1980s and 1990s by enormous anthropogenic inputs. Aerosol samples collected concurrently with our seawater samples support our determination that Pb in the surface waters is partially of natural origin (Shelley et al., 2017), and work by Prospero 25 et al (2012) introduces the possibility that much of the dust in the Irminger Sea and Icelandic Basins is actually from high latitude sources such at Icelandic dust. Future work to better constrain end-members could validate this hypothesis. Pérault, and Emmanuel de Saint-Léger. GEOVIDE nutrient data was obtained by Manon Le Goff, Emilie Grossteffan, Morgane Gallinari, and Paul Tréguer. We also thank the officers and crews of the RV Atlantis II (1989) Biogeosciences Discuss., https://doi.org /10.5194/bg-2018-29 Manuscript under review for journal Biogeosciences Discussion started: 2 February 2018 c Author(s) 2018. CC BY 4.0 License. Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-29 Manuscript under review for journal Biogeosciences Discussion started: 2 February 2018 c Author(s) 2018. CC BY 4.0 License. Transect, Deep-Sea Research II, 116, 208-225, doi:10.1016/j.dsr2.2014.11.011 2015 Planquette, H. and Sherrell, R.M.: Sampling for particulate trace element determination using water sampling bottles: methodology and comparison to in situ pumps, Limnol. Oceanogr. Methods, 10, 5, 367-388, doi:10.4319/lom.2012.10.367, 2012. Biogeosciences Discuss., https://doi.org/10.5194/bg-2018 Manuscript under review for journal Biogeosciences Discussion started: 2 February 2018 c Author(s) 2018. CC BY 4.0 License.