River fluxes to the sea from the ocean’s 10Be/9Be ratio Friedhelm von Blanckenburg, Julien Bouchez

To cite this version:

Friedhelm von Blanckenburg, Julien Bouchez. River fluxes to the sea from the ocean’s 10Be/9Be ratio. Earth and Planetary Science Letters, Elsevier, 2014, 387, pp.34-43. ￿10.1016/j.epsl.2013.11.004￿. ￿hal- 02133437￿

HAL Id: hal-02133437 https://hal.archives-ouvertes.fr/hal-02133437 Submitted on 18 May 2019

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Originally published as:

von Blanckenburg, F., Bouchez, J. (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio. - Earth and Planetary Science Letters, 387, 1, 34-43

DOI: 10.1016/j.epsl.2013.11.004 River fluxes to the sea from the ocean’s 10Be/9Be ratio

Friedhelm von Blanckenburga, * and Julien Bouchezb aEarth Surface Geochemistry, Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, Telegraphenberg, 14473 Potsdam, Germany

*Communicating author. [email protected]. Also at: Institute of Geological Sciences, Freie Universität Berlin, Germany. c present address: Institut de Physique du Globe de Paris, 1 rue Jussieu, 75232 Paris 05, France

Earth Planet Sci Letters 2014, 387:34-43; doi 10.1016/j.epsl.2013.11.004

Abstract

The ratio of the meteoric cosmogenic radionuclide 10Be to the stable isotope 9Be is proposed here to be a flux proxy of terrigenous input into the oceans. The ocean’s dissolved 10Be/9Be is set by (1) the flux of meteoric 10Be produced in the atmosphere; (2) the denudational flux of the rivers discharging into a given ocean basin; (3) the fraction of 9Be that is released from primary minerals during weathering (meaning the 9Be transported by rivers in either the dissolved form or adsorbed onto sedimentary particles and incorporated into secondary oxides); and (4) the fraction of riverine 10Be and 9Be actually released into seawater. Using published 10Be/9Be data of rivers for which independent denudation rate estimates exist we first find that the global average fraction of 9Be released during weathering into river waters and their particulate load is 20% and does not depend on denudation rate. We then evaluate this quantitative proxy for terrigenous inputs by using published dissolved seawater Be isotope data and a compilation of global river loads. We find that the measured global average oceanic dissolved 10Be/9Be ratio of about 0.9×10-7 is satisfied by the mass balance if only about 6% of the dissolved and adsorbed riverine Be is eventually released to the open ocean after escaping the coastal zone. When we establish this mass balance for individual ocean basins good agreement results between 10Be/9Be ratios predicted from known river basin denudation rates and measured ocean 10Be/9Be ratios. Only in the South Atlantic and the South Pacific the 10Be/9Be ratio is dominated by advected Be and in these basins the ratio is a proxy for ocean circulation. As the seawater 10Be/9Be ratio is faithfully recorded in marine chemical precipitates the 10Be/9Be ratio extracted from authigenic sediments can now serve to estimate relative changes in terrigenous input into the oceans back through time on a global and on an ocean basin scale.

Keywords: Seawater chemistry; cosmogenic isotopes; weathering; erosion; river flux

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 1

1. Introduction

10 9 10 The Be/ Be isotope ratio combines the meteoric cosmogenic radionuclide Be (T1/2 = 1.39 Myr), deposited onto the Earth's surface by precipitation, with its stable counterpart 9Be, released from rocks by weathering. Given this combination, the 10Be/9Be ratio in seawater is sensitive to continental denudation rates. The average residence time of Be in seawater being only ca. 600-1000 years, the ratio differs between ocean basins (e.g. Bourlès et al., 1989; Brown et al., 1992; Frank et al., 2009; Kusakabe et al., 1990; von Blanckenburg et al., 1996). Correspondingly, dissolved 10Be/9Be ratios in seawater are lowest in the Mediterranean Sea, a small ocean basin dominated by high terrigenous input and limited exchange with the global ocean; and highest in the Pacific, which collects a large flux of meteoric 10Be over its vast surface area while receiving comparatively low terrigenous input given the size of the ocean basin. It has been proposed that the 10Be/9Be ratio extracted from authigenic marine sediments, recording the surrounding seawaters’ 10Be/9Be ratio, traces continental denudation back through time (Bourlès et al., 1989; von Blanckenburg and O'Nions, 1999; Willenbring and von Blanckenburg, 2010a). To date only very coarse (My) time resolution records of paleo-denudation over the last 10 My were reconstructed (Willenbring and von Blanckenburg, 2010a). When applied at higher time resolution the method will allow to evaluate changes in denudation rate as a function of Quaternary climate shifts. Ultimately, the temporal resolution is limited only by the residence time of Be in the respective ocean basin, and by the resolution of the sedimentary record.

A geochemical framework suited to explore how erosion and weathering set the 10Be/9Be ratio of soils and rivers was suggested recently by von Blanckenburg et al. (2012). A river drainage basin’s denudation rate can be calculated by measuring the 10Be/9Be ratio of the dissolved Be or reactive Be carried by rivers (meaning adsorbed or bound to secondary oxyhydroxide precipitates) provided that the flux of meteoric 10Be into this basin and the concentration of 9Be in bedrock are known, and that the fraction of 9Be released from primary minerals during weathering can be estimated. The 9Be thus released and the meteoric 10Be deposited onto the continents are delivered through riverine transport to the oceans, where they mix with 10Be deposited directly onto the ocean surface. Aeolian input only plays a minor role for the modern oceanic 9Be budget (von Blanckenburg et al., 1996; Willenbring and von Blanckenburg, 2010a). After entry into the ocean, particle-reactive Be can be captured in estuaries and shelf areas (Brown et al., 1992; Kusakabe et al., 1991) and at ocean margins (Anderson et al., 1990; Lao et al., 1992). Yet these sites of scavenging are also sites of release of reactive metals from shelf and slope sediments during diagenetic reactions and pore fluid expulsion (Arsouze et al., 2009; Jeandel et al., 2011). Dissolved Be that reaches the open ocean is entrained by the gyres of the major ocean basins that have the remarkable ability to homogenise this isotope ratio despite the reactive nature of Be and the high scavenging intensity at von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 2 the ocean margins (Igel and von Blanckenburg, 1999). For example, in the Pacific Ocean the Be is mixed to such an extent that the 10Be/9Be ratio exhibits a 13% dispersion only, calculated from measured surface water ratios, despite input 10Be/9Be ratios spanning a range of up to five orders of magnitude (von Blanckenburg and Igel, 1999). At the scale of an ocean basin, Be with a uniform 10Be/9Be ratio is then delivered to deep waters, where 9Be is present at relatively invariant concentrations of 30 pM (Kusakabe et al., 1990). Given that the deep ocean residence time is about 600-1000 years (Kusakabe et al., 1990; von Blanckenburg et al., 1996) this Be compartment is characterized by 10Be/9Be ratios that are distinct between ocean basins but that are more or less uniform within a basin (Kusakabe et al., 1990).

Considering this proxy’s considerable potential as tracer for terrigenous erosion and weathering, it is timely to develop means for the translation of present and past 10Be/9Be ratios to global estimates of the transfer of terrigenous matter into the oceans. The aim of this contribution is therefore 1) to determine, using global river denudation and oceanic 10Be/9Be data, the terrigenous input fluxes of these two isotopes into the global oceans, as well as into individual ocean basins; 2) to provide estimates of the fractional release of this metal into the dissolved form once delivered into the oceans; 3) to explore the fidelity of the oceanic 10Be/9Be ratio as a quantitative proxy for past terrigenous inputs to the ocean when extracted from authigenic marine sediments. We do this by combining published dissolved 10Be/9Be data from all ocean basins with estimates of the terrigenous fluxes of Be into these basins.

2. The oceanic Be isotope mass balance

The first aim of this study is to characterise the 10Be/9Be ratio established in the oceans following transfer of terrigenous 9Be and 10Be via rivers and the addition of meteoric 10Be deposited directly onto the ocean’s surface (Figure 1). We follow the 10 9 terminology of von Blanckenburg et al. (2012): ( Be/ Be)diss is the isotope ratio of Be 10 9 dissolved in river waters (Bediss); and ( Be/ Be)reac is the isotope ratio of reactive Be

(Bereac), adsorbed onto the particulate load of rivers and contained in Be-rich secondary phases, such as Fe-Mn oxihydroxides. Provided the contact time between 10 9 10 9 river water and river particles is sufficiently long, ( Be/ Be)reac and ( Be/ Be)diss equilibrate during riverine transport to a common ratio as shown previously (von Blanckenburg et al., 2012). In a given river and at steady-state, this ratio is:

10  10 Be   10 Be  F Be   =   = riv (1)  9   9  9 9 Be 9 Be  Be reac  Be diss D ⋅ [ Be]parent ⋅ (freac + fdiss ) von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 3

10 Be 10 where Friv describes the atmospheric flux of meteoric Be onto the river basin (in -2 -1 10 Be atoms m yr ; note that in von Blanckenburg et al. (2012) we call this flux Fmet ), D the river basin’s denudation rate that comprises both physical erosion and chemical -2 -1 9 9 weathering (in kg m yr ), [ Be]parent is the mean Be concentration of the river -1 -1 9Be 9 Be 9 basin’s parent rock (in atoms kg or mol kg ), and freac + f diss is the fraction of Be released during weathering from the parent rock, meaning the fraction of 9Be that is 10 9Be 9 Be ultimately available for mixing with dissolved Be. Hereafter we call freac + f diss the 9 9 mobile fraction of Be. We assume for simplicity that the remainder of the Be, which is locked in primary minerals, does not enter the oceanic Be cycle. Therefore, the terrigenous Be finally available for release and mixing in seawater is supplied either in the reactive form (Bereac) as borne by the solid erosional load of rivers or in the dissolved form (Bediss) in river water. However, not all mobile Be is ultimately delivered to the open ocean, as in the coastal ocean reactive Be might be deposited with the particles that carry this Be. The fraction of reactive Be thus trapped might differ from the fraction of reactive Be carried by rivers, as reactive and dissolved Be can exchange upon entry in the oceans. We represent the efficiency of overall Be delivery to the open ocean by a fractional factor φdel that is identical for both isotopes.

Essentially, φdel quantifies the fraction of riverine mobile Be surviving the sink provided by the estuaries and the coastal interface. Once in dissolved form in the open ocean, all Be from the continents is rapidly homogenised with the meteoric 10Be deposited directly onto the surface oceans. The average dissolved 10Be/9Be ratio of 10 9 an ocean basin, termed ( Be/ Be)oc, can be calculated as follows:

10  10 Be  A ⋅ F Be + A ⋅(E ⋅φ ⋅[10 Be] + Q ⋅φ ⋅[10 Be] )   = oc oc riv riv del reac riv del diss (2)  9  ⋅ ⋅φ ⋅ 9 + ⋅φ ⋅ 9  Be oc Ariv (Eriv del [ Be]reac Qriv del [ Be]diss )

10 Be 10 where Foc is the flux of meteoric Be deposited onto the ocean surface (in atoms -2 -1 2 m y ) averaged over Aoc, the surface area of the ocean (in m ), Eriv is the average -2 -1 erosion rate (in kg m y ) for the river basins discharging into the ocean, Ariv is the 2 total surface area of these basins (in m ), and Qriv is the total runoff of these rivers (in -1 10 9 10 9 m s ). [ Be]reac, and [ Be]reac are the flux-weighted average reactive Be and Be 10 9 concentrations in the river sediments, while [ Be]diss, and [ Be]diss are the flux- weighted average dissolved 10Be and 9Be concentrations of rivers draining to the ocean. At steady-state, the amount of meteoric 10Be delivered by rivers to the ocean is equal to that deposited onto the continental surface drained by those rivers:

10 10 10 Be Ariv ⋅(Eriv ⋅[ Be]reac + Qriv ⋅[ Be]diss )= Ariv ⋅ Friv (3)

10 Be 10 where Friv is averaged over Ariv. The steady-state flux of this continental Be to the ocean does not depend on denudation rate. First, the amount of in situ-produced 10Be carried by solid river loads is negligible as compared to the amount of meteoric 10Be. Second, if steady state between sediment production and sediment export is indeed attained, large-scale storage of sediment in continental alluvial deposits does von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 4

10 not result in net trapping of Be. Hence, regardless of Eriv and Qriv and the amount of sediment stored in alluvial deposits, eventually all 10Be precipitated is quantitatively transported to the oceans. However, if the residence time during alluvial storage is significant relative to the half-life of 10Be (1.39 Myr), some of the 10Be might decay before reaching the ocean. We ignore the effect of this radioactive decay as it amounts to a maximum error of only 2% onto the ocean 10Be/9Be ratio if all terrigenous 10Be had decayed before reaching the ocean (corresponding to the 10 maximum proportion of ocean Be supplied by rivers, which is Ariv/(Ariv+Aoc)× φdel).

Similarly for 9Be:

9 9 A ⋅ E ⋅ 9Be + Q ⋅ 9Be = A ⋅ D ⋅ 9Be ⋅ f Be + f Be (4) riv ( riv [ ]reac riv [ ]diss) riv riv [ ]parent ( reac diss )

where Driv is the average denudation rate in the basin drained by rivers flowing to the 9Be 9 Be 9 ocean, and freac + f diss , the mobile fraction of Be, is averaged over these river basins. Substituting equations (3) and (4) in equation (2) the final equation describing 10 9 ( Be/ Be)oc, conceptually illustrated in Figure 1, is:

 A  10 10  oc ⋅ F Be +φ ⋅ F Be 10   oc del riv  Be   Ariv    = (5)  9  9 9 Be 9 Be  Be oc φdel ⋅ Driv ⋅[ Be]parent ⋅(f reac + f diss )

Equation (5) can also be applied to large endorheic river basins, with Aoc then representing the lake area. If the lake area is small relative to the river basin area

(Aoc << Ariv), then equation (5) simplifies to equation (1). When applied to a large endorheic basin, the authigenic phase of sediment precipitated into the lake would 9 9 record delivery of Bereac and Bediss into the lake (equation (5)).

3. The modern ocean budget of Be

3.1 Data sources

10 9 We compare modelled ( Be/ Be)oc calculated from published estimates of the source fluxes as defined in equation (5) with 10Be/9Be ratios published for dissolved Be in modern seawater. We do not use 10Be/9Be ratios measured on the surface scrapings of hydrogenetic Fe-Mn crusts (e.g. (von Blanckenburg et al., 1996)) given that these integrate over several 105 years, a timescale over which the terrigenous fluxes of 9Be might have changed relative to the modern riverine fluxes used here. Seawater Be isotope concentrations were measured in the North Pacific (Kusakabe et al., 1987; Xu, 1994), in the Atlantic (Measures et al., 1996), in the Mediterranean Sea (Brown et al., 1992), and in the Arctic Ocean (Frank et al., 2009). (Table S1, Supplement S1). 10Be and 9Be concentrations are strongly depth-dependent and follow a nutrient- von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 5 type depth-profile in most ocean basins (Kusakabe et al., 1990), which means that concentrations of dissolved Be are low at the surface after uptake onto particles and increase with depth as Be is released from sinking particles. Although the 10Be/9Be ratio is not depth-dependent to the same extent, it also varies with depth, given that most 10Be is delivered to the surface ocean through direct precipitation, whereas the deep water 10Be/9Be ratio is influenced by advection of Be via the thermohaline circulation. Therefore, we group our data into “shallow water”, comprising all measurements up to 1000 m depth, essentially thermocline water, and “deep water”, comprising all measurements between 1000 m water depth and the seafloor. No dissolved Be has been measured to date in the Indian Ocean. We instead use the 10Be/9Be ratio measured on leachates of deep ocean surface sediments from the Indian Ocean (Bourlès et al., 1989), assumed to represent the isotopic composition of modern bottom waters. The ocean basin averages are summarised in Table 1. Weighting the Be isotope data by ocean basin area, we obtain a global ocean 10Be/9Be ratio of 0.97 ± 0.20 × 10-7 for shallow water, and 0.84 ± 0.08 × 10-7 for deep water, respectively.

10 9 10 9 We next compare these measured oceanic Be/ Be ratios with ( Be/ Be)oc predicted 10 Be 10 Be by equation (5). We use the following fluxes. Foc and Friv are based on the global flux distribution of Willenbring and von Blanckenburg (2010b). This map is an average of the model distributions maps of Field et al. (2006) and Heikkilä et al. 10 Be (2008). We extract an ocean basin-specific Foc and, for simplicity, use the same value for the flux of meteoric 10Be onto the river basins discharging into that ocean 10 Be basin Friv given that the meteoric flux varies mainly as a function of latitude and thus is similar between the oceans and the adjacent continental area (Table 1). The flux of riverine material to the ocean is more than 20 times greater than the atmospheric dust flux, and it is likely that the former dominates the marine chemistry of beryllium, similar to numerous other elements dissolved in seawater (Oelkers et al., 2012). We compiled the denudational flux of solid and dissolved matter discharged into each ocean basin from the global river flux database of Milliman and Farnsworth (2011) (Supplement S2). The data covers a global exhoreic basin area 6 2 (global Ariv) of 84.3 × 10 km , which is lower than the estimate of Peucker- Ehrenbrink (2009). We calculate the denudation rate D for continental areas draining into each ocean basin and for the global ocean by summing up all suspended and dissolved loads of the individual rivers draining into a particular basin, and by dividing the obtained flux by the corresponding drainage area. We obtain a global sediment yield of 147 t km-2 y-1 and a global dissolved yield of 36 t km-2 y-1, corresponding to an annual global sediment delivery to the oceans of 12.4 × 109 t and a global dissolved flux of 3 × 109 t. Correcting the dissolved flux for cyclic salts would change the estimates of Driv by only a small amount, given the predominance of the particulate over the dissolved flux. The surface areas used and fluxes calculated are summarised in Table 1. As no modern ocean 10Be/9Be data is available for the South Pacific, we present the global denudation estimate both including and excluding the

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 6

South Pacific. The difference introduced into the global denudation flux is small as

Ariv in the South Pacific is only 10% of that of the North Pacific (Table 1). We confirm previous analyses (von Blanckenburg et al., 1996; Willenbring and von Blanckenburg, 2010a) that the Be input through dust into the oceans is most likely lower than a few percent of the total 9Be input (Supplement S3). Finally, for 9 calculating input fluxes, we used [ Be]parent of 2.5 ppm as a reasonable estimate of the mean land surface lithologies (von Blanckenburg et al., 2012). In order to predict 10 9 a global ( Be/ Be)oc ratio using equation 5, we are left with two unknowns: the mobile 9 9Be 9 Be fraction of Be freac + f diss and the fractional delivery of mobile Be from the ocean margins into the open ocean, φdel.

3.2 The mobile fraction of 9Be

We use the river Be isotope data base from von Blanckenburg et al. (2012) and solve 9 9Be 9 Be equation (1) for each river for the mobile fraction of Be freac + f diss (Figure 2). The set of rivers used in this calculation is representative of a large variety of lithologic, climatic, and tectonic regimes. When excluding the rivers draining the highly weathered cratonic rocks (most likely over-represented in our data set by several rivers draining the South American shields), the global rivers appear to share a common mobile fraction of 9Be of 0.18 ± 0.08. When including those rivers draining shield rocks, we obtain an average mobile fraction of 9Be of 0.24 ± 0.18. We use a rounded global mobile fraction of 9Be of 0.2, meaning that 20% of the Be contained within the rocks is liberated by weathering. Given that in granitoid rocks Be is contained in highly weatherable minerals such as plagioclase and biotite (Grew, 2002), this value might appear surprisingly low. However, on a global scale, silicate lithologies are dominated by clastic sedimentary rocks (Amiotte-Suchet et al., 2003), in which Be will be contained in minerals such as muscovite and illite, that readily survive continental weathering.

3.3 The fractional delivery of terrigenous Be into the open ocean

As the behaviour of Be in estuarine and shelf areas is not yet fully understood, no a- priori estimate is possible for the fractional delivery factor φdel. We thus first solve 9 equation (5) for the global value of φdel by using a mobile fraction of Be of 0.2 in conjunction with the global denudational fluxes, meteoric 10Be fluxes, and areas described in the previous section, and later compare the calculated global φdel value with those reported for other reactive metals. A global deep ocean (excluding the 10 9 -7 South Pacific for lack of ocean Be data) model ( Be/ Be)oc of 0.88 × 10 , similar to the measured ratio (Table 1), is obtained if φdel has a value of 0.063 ± 0.036. The uncertainty of this estimate is presented in Appendix 1. This φdel value means that 6.3% of both 10Be and the 9Be entering the oceans from the continents, either in the dissolved or in the reactive form, are eventually released into the open ocean gyres in the dissolved form, where they mix with the additional 10Be deposited onto the oceans via direct precipitation. φdel does not describe the processes that scavenge von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 7

10Be precipitated directly into the open oceans, termed boundary scavenging processes (Anderson et al., 1990; Lao et al., 1992). Instead φdel quantifies removal of Be at those sites at which the 10Be and 9Be isotopes derived from the continents are not yet homogenised with the 10Be originating from direct precipitation into the oceans. Where the oceans 10Be/9Be ratio is already homogenised (Igel and von Blanckenburg, 1999) boundary scavenging of Be does not affect the oceans 10Be/9Be ratio.

We can now compare the derived global delivery factor φdel with previous studies on other metals. Commonly, the delivery factor to the ocean has been reported relative to the total riverine (i.e. including primary minerals, secondary precipitates, and dissolved) elemental flux to the ocean, whereas φdel in our study quantifies the fractional delivery of dissolved and reactive Be only. From our results, a total riverine 9 Be 9 Be Be global delivery factor of 1.2% can be calculated from φdel × (f reac + f diss ). This value would not change even if primary minerals released some of their 9Be in the 9 9 Be 9 Be seawater. Oceanic release of Be from primary minerals would shift (f reac + f diss ) in the mass balance to higher values at the cost of reduced φdel. The 1.2% value compares well with the global delivery factor estimated for the rare earth element neodymium, a similarly reactive element, for which 3-5% transfer to the open ocean is calculated (Arsouze et al., 2009). Rempfer et al. (2011) found that 6% of the ocean Nd is supplied via direct river discharge. About 10% of the riverine iron flux, another particle-reactive element, survives estuarine loss (Raiswell, 2006). 3-5% of Th, a very particle-reactive element, was estimated to be dissolved from continental input (Roy- Barman et al., 2002).

3.4 The riverine flux into individual ocean basins traced by their 10Be/9Be ratio

The accuracy of our global ocean mass balance cannot be fully evaluated as φdel remains an adjustable variable. However, assuming that the global delivery factor 9 φdel and the uniform mobile fraction of Be (0.2) across different rivers are valid for each separate ocean basin, we can test our approach for internal consistency by applying the model expressed by equation (5) on a basin-by-basin case. In this case the widely varying denudation rates contributing to the individual basins as well as 10 9 their range of Ariv/Aoc ratios should be reflected in the individual basins’ ( Be/ Be)oc.

The assignment of the associated continental drainage areas is documented in Supplement S2. To perform this test we separate the global ocean into the North Pacific basin including the South China and the Bering Sea, the South Pacific (south of the equator), the North Atlantic including the North Sea and the Caribbean, the South Atlantic (separated from the N Atlantic at 5° S, as the surface ocean circulation system results in a distinct 10Be/9Be domain south of this latitude, (McDonald and Wunsch, 1996)), the Mediterranean (excluding the Black Sea, as we assume minimal exchange of Be between these two basins), the Indian Ocean including the Mallaca Strait, North-Eastern Australia, and the seas of South-East Asia which are assumed von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 8 to deliver most of the terrigenous input into the Indian Ocean via the Indonesian throughflow. We summed up Ariv for the individual drainage basins of Milliman & Farnsworth (2011). The results of the model are given in Table 1, and the comparison to measured shallow and deep water 10Be/9Be ratios is shown in Figure 3.

10 9 The model predicts low ( Be/ Be)oc values for the North Atlantic and the Mediterranean Sea reflecting their high denudational input, and the Arctic ocean, 10Be reflecting its small Aoc and low F . For the Indian Ocean, despite its high inputs metoc caused by a high denudation rate of 400 t km-2 y-1, only intermediate model 10 9 10 ( Be/ Be)oc values are predicted given that its large Aoc also results in a high Be 10 9 input. The North Pacific displays the highest model ( Be/ Be)oc ratios. Only the South Pacific, for which no measured Be data exists, is predicted to feature an even higher 10 9 -7 ( Be/ Be)oc of 9.8 × 10 , given its very small Ariv/Aoc ratio and hence its minor terrigenous input (Table 1).

9 Using the global estimates for φdel of 0.063 and a mobile fraction of Be of 0.2 we obtain excellent agreement between modeled and measured 10Be/9Be seawater ratios for the North Pacific, the Indian, the North Atlantic, the Arctic Ocean and the Mediterranean Sea for both shallow and deep water (Figure 3). Only the South Atlantic does not match the predictions. There, the model predicts much higher 10 9 ( Be/ Be)oc than observed in either shallow or deep water, although the agreement is slightly better for shallow water. This misfit can be explained by the low denudation rate in basins surrounding the southern Atlantic and the small associated Ariv, which makes the seawater isotope composition extremely sensitive to lateral inputs through ocean circulation. We quantify the effect of advection of deep and shallow water by means of a simple flux-based steady-state mixing model (Supplement S4). For each ocean basin we take the river-sourced fluxes of 10Be and 9Be and the meteoric flux 10 Be Foc based on the parameters of Table 1. To these we add advective fluxes based on water mass transport between basins (Ganachaud and Wunsch, 2000), and on the measured 9Be concentrations and 10Be/9Be ratios of Be dissolved in water. We then calculate ratios for the mixture between the denudation and advective inputs 10 9 ( Be/ Be)mixture (Figure 5). The deep water of the South Atlantic is essentially a mixture of North Atlantic Deep Water and Antarctic Bottom and Intermediate Water. Therefore, its isotope ratio is dominated by those of the northern Atlantic (0.6 × 10-7) -7 10 9 and the Pacific (1 × 10 ). The South Atlantic ( Be/ Be)mixture is indeed closer to the measured value (Figure 3). We can expect a similar situation for the Southern Pacific, for which no dissolved Be data exists. 10Be/9Be ratios from surface scrapings of Fe-Mn crusts (von Blanckenburg et al., 1996) are actually around 1 × 10-7, much 10 9 -7 lower than the prediction of the denudation-based ( Be/ Be)oc, 9 × 10 (Table 1), and similar to the composition of advected circumpolar water.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 9

4. Exploring past changes in terrigenous weathering and erosion using sedimentary 10Be/9Be ratios

10 9 To demonstrate that Be/ Be ratios from the authigenic phase of sediment can trace terrigenous inputs back through time we first discuss the ocean input processes that the 10Be/9Be ratio actually quantifies, then evaluate whether this signal is transmitted unmodified through the coastal trap, and finally assess its potential to record changes in denudation rates.

4.1 The sedimentary 10Be/9Be ratio: a proxy for denudation, for weathering, or for Be delivery through the coastal zone?

For 10Be/9Be ratios to faithfully record denudation rates, the first condition is that a roughly uniform and constant mobile 9Be fraction is delivered by rivers into the 9Be 9 Be coastal zone, meaning that at the ocean basin scale freac + f diss does not change with D. For the large river scale we have already shown that this condition is fulfilled (see section 3.2), as most global rivers share a common mobile fraction of 9Be of 0.2. This observation is further discussed in section 4.2. We have also shown previously 10 9 10 9 that in large rivers ( Be/ Be)reac and ( Be/ Be)diss likely equilibrate to a roughly 10 9 similar ratio (von Blanckenburg et al., 2012). Hence globally both ( Be/ Be)reac and 10 9 ( Be/ Be)diss of Be delivered to the coastal zone essentially record D. For ocean 10Be/9Be ratios to faithfully record fluxes of continental inputs to the sea, another requirement is that φdel does not depend on those inputs. We have evidence that this condition is fulfilled between the different ocean basins as variations of 10Be/9Be ratios between ocean basins primarily reflect differences in their denudational fluxes and not in φdel (section 3.4).

These observations suggest that Be behaviour in rivers and coastal ocean is uniform and independent on D. We invoke two different scenarios that can be regarded as end member pathways for Be delivery to the open ocean to account for these observations. Importantly, these scenarios have different implications for the interpretation of sedimentary 10Be/9Be ratios in terms of past continental input to the seas.

A first possibility is that only dissolved Be from the continents eventually enters the open ocean. Such a coastal bypass process has been shown for some of the rivers draining into the Arctic Ocean (Frank et al., 2009). In this case the fraction of total 9 9 Be 9 Be 9 Be Be delivered φdel × ( f reac + f diss ) is essentially identical to fdiss in rivers. This scenario is not entirely unlikely as from our results, a total riverine Be delivery fraction 9 Be of 1.2% was calculated (section 3.3) which is comparable with the values for fdiss of 2-3% measured in the Amazon river (Table 2 in von Blanckenburg et al. (2012)). In this case the ocean 10Be/9Be ratio directly reflects the chemical weathering flux of 9Be into the oceans, i.e. the release flux of 9Be from primary minerals minus the uptake 9 Be 9 Be 9 Be flux into the reactive fraction during weathering. At constant f reac + f diss and fdiss ,

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 10 the ocean 10Be/9Be ratio then reflects the ocean basin-wide denudation flux D. One prediction from this scenario is that changes in continental weathering back through time should be instantaneously transmitted into shifts in the ocean 10Be/9Be ratio. We know of such a behaviour for Nd, of which the radiogenic isotope ratios change abruptly over glacial - interglacial time scales (Burton and Vance, 2000; Gourlan et al., 2010; Wilson et al., 2012).

A perhaps more realistic version of this first scenario is that the partitioning between

Bediss and Bereac occurs to some extent through adsorption-desorption processes in rivers and coastal ocean. Therefore, the value of φdel should be a function of this partitioning, which in turn depends on particle density and the particle-water partition Be coefficient K d (Supplement S5). One might suspect that through changing sediment delivery to the coastal ocean, φdel is a function of D itself. However, global river data show that river particle concentration and D are not correlated (Milliman and Farnsworth, 2011). Furthermore, even if D and particle concentration in the coastal Be ocean were positively correlated, K d itself depends in an inverse way on particle concentration, an effect known as “particle concentration effect” (Hawley et al., 1986). These two effects largely compensate each other (Supplement S5, see also sensitivity analysis below). Hence the ocean 10Be/9Be ratio still reflects the ocean basin-wide weathering flux.

The second scenario features co-precipitation of Be as a significant removal process in rivers and in the coastal ocean. Such co-precipitation might occur if Be is, for example, quantitatively scavenged with Fe(III)oxyhydroxides upon entry in the coastal ocean. Studies on estuarine scavenging of 9Be for example used flux balances in which it was assumed that over short time scales, scavenging in estuaries and at ocean boundaries removes riverine 9Be almost quantitatively, given the highly reactive nature of Be at near-neutral pH (Kusakabe et al., 1990; Kusakabe et al., 1991). However, over longer time scales, we expect that some of the scavenged Be is released. Early diagenetic processes on the shelves, for example those coupled to the iron redox cycle, can result in fast precipitation-dissolution 9 cycles over which a minute fraction of Bereac leaks into seawater as Bediss. Such scavenging and release processes are similar to what has been termed boundary exchange (Arsouze et al., 2009; Lacan and Jeandel, 2005; Wilson et al., 2012): the 10 9 extent of this Be leak sets the value of φdel. In this case ocean Be/ Be ratios reflect the flux of mobile (reactive plus dissolved) 9Be delivered by rivers to the coastal ocean. As mobile 9Be is dominated by sediment-borne reactive Be (Table 2 in von Blanckenburg et al. (2012)), 10Be/9Be ratios reflect the ocean basin-wide denudation flux D as transmitted by the sedimentary flux into the coastal ocean. However, this signal is likely to be damped in the past 10Be/9Be record, depending on the time scale of release of scavenged Be. This time scale is not straightforward to estimate, ranging from years if pore fluid expulsion from coastal sediments into the overlying water column is the major Be release process, to kiloyears if release is dominated by shelf instability during sealevel lowstands. However, significant variations in the Nd von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 11 isotopes record over glacial-interglacial cycles suggest that such a damping, if any, does not completely smooth the source signal.

4.2. Precision of paleo-D estimates

10 9 We have seen that for individual ocean basins the modern ( Be/ Be)oc can be predicted from independent estimates of D to within a factor of two (Figure 3). We envisage this accuracy as applicable only to the determination of absolute rates of paleo-D from past 10Be/9Be ratios as some of the mismatch might be caused by inaccurate estimates of the suspended and dissolved river fluxes used, in addition to inaccuracies inherent to the proxy itself. Our ability to reconstruct past relative fluctuations in D in a given ocean basin is not impaired by these systematic inaccuracies, but rather requires examination of the potential sources of error inherent to the 10Be/9Be proxy, that we now present.

First, some of this bias might be introduced through ocean circulation, which is quantifyable and hence correctable. This correction was done in a very simple way in section 3.4 but can potentially be performed more accurately by use of general circulation models including tracer transport (e.g.Rempfer et al., 2011).

More importantly the precision of paleo-D is limited by the precision at which we 10 know the paleo-10Be production F Be , which in turn depends on variations in the geomagnetic field strength (Carcaillet et al., 2004), as we assume that shorter-term fluctuations caused by changes in solar modulation are averaged out (Willenbring 10 and von Blanckenburg, 2010b). In principle, variations in paleo- F Be can be corrected using geomagnetic paleo-intensity records. However, large discrepancies in both amplitude and temporal variability have been reported between different geomagnetic paleo-intensity records (Channell et al., 2009; Valet et al., 2005; Ziegler et al., 2011). Moreover, unexplained discrepancies exist when comparing marine sediment 10Be records with different paleo-intensity records (Christl et al., 2010; Frank et al., 1997; Knudsen et al., 2008). For the case that past 10Be/9Be ratios can be determined on sedimentary sequences with extremely good age control, and if 10 variations in paleo- F Be are also known accurately for this time interval, the error 10 introduced by F Be will be small. In all other cases, this error depends on both the integration time of the sediment sample and on the potential offset between the age models of the 10Be/9Be and geomagnetic paleo-intensity records. We use the paleo- 10 F Be record of Christl et al. (2010) to quantify this error (Figure 5). We find that at low 10 integration time (a few ky) the possible deviation from the actual F Be amounts to 3 to 17%, depending on the offset between the age models. At high integration time 10 scale (> 50 ky) errors are small and amount to 2 to 5%, as most fluctuations in F Be are averaged out (Figure 5).

The fidelity of the proxy further depends on the condition that the parameters used to convert ocean 10Be/9Be ratios into D are not themselves strong functions of D. These

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 12

9 Be 9 Be parameters are Ariv and Aoc, φdel and f reac + f diss . The ratio Aoc/Ariv in equation (5) shifts from 4.3 in interglacials to 3.4 at low-sealevel in glacials (François et al., 1998). 10 9 This shift results in a small uniform offset in the relationship between ( Be/ Be)oc and D amounting to an offset of ca. 20 t km-2 y-1 in D (Figure 6) relative to the modern global average (Table 1). This offset can be corrected for by using the global sealevel curve. The maximum potential dependence of φdel on D can be expressed as a function of sediment delivery to the coastal ocean (see Supplement S5), where sediment delivery is assumed to vary proportionally to D. As explained in section 4.1, during times of high D increase in particle concentration is compensated by the resulting lowering of the partition coefficient of Be between the reactive and dissolved phases (following Hawley et al.,(1986)). The difference between the constant- and D -2 -1 dependent-φdel scenarios results roughly in an offset D of roughly ± 20 t km y . For 9Be 9 Be the dependence of freac + f diss on D, an upper extreme case in the form of a power law fit through the upper data envelope of rivers shown in Figure 2 can be calculated, which is likely an overestimate. Errors of ± 20 to ± 100 t km-2 y-1 result. Using a lower 10 9 -7 and an upper bound for the measured oceans Be/ Be ratio, (0.7 to 1.5 × 10 , Table 10 Be 1), and assuming that the sources of errors (in F , Ariv and Aoc, φdel, and 9 Be 9 Be f reac + f diss ) do not compound each other, an overall uncertainty of ca. 20-30% can be estimated on past relative variations in D (for D being similar to the modern value).

9 Finally, changes in the relative contribution of lithologies with different [ Be]parent can affect the oceanic 10Be/9Be. Li and Elderfield (2013) have recently suggested that weathering from volcanic islands decreased in the late Neogene, whereas weathering of felsic continental rocks simultaneously increased. Keeping a given ocean basins 10Be/9Be ratio constant over the same period (Willenbring and von Blanckenburg, 2010a) would require the fortuitous case that the product of D × 9 9 Be 9 Be [ Be]parent × ( freac + fdiss ) × φdel resulting from the terrigenous input into that basin 9 remained constant. As [ Be]parent is 5 times lower in basalts than in continental felsic rocks (von Blanckenburg et al., 2012), this condition is only met if 9Be in basalt is completely released from primary volcanic rocks during subaerial weathering or in 9 Be 9 Be seawater (Oelkers et al., 2012). In this case ( freac + fdiss ) is 1 and the product 9 9 Be 9 Be [ Be]parent × ( freac + fdiss ) is identical between basaltic and felsic settings. Unless an equivalent change in φdel occurs between lithologies, variations in D or their absence are indeed reflected in the oceanic 10Be/9Be ratios. However, the near-neutral pH of 9 rivers draining basaltic lithologies likely results in exceedingly low riverine [ Be]diss, and, once released from primary minerals, all Be will be carried in the reactive form. For basalt weathering to influence the oceanic 10Be/9Be ratio at all, a further requirement is hence that Bereac is released in seawater. This might not be the case if all reactive Be is locked into secondary clays. In that case the oceanic 10Be/9Be ratio records paleo-D of felsic continental rocks only.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 13

5. Potential Applications

The advantage of this new 10Be/9Be proxy lies in its ability to quantitatively determine relative variations in past denudation rate from the authigenic phase of marine sediment. In this regard the proxy provides fundamental advantages over the commonly used radiogenic or stable isotope proxies that do not provide any direct information on fluxes. These estimates can be determined on a range of temporal resolution, from a minimum of 600 y (residence time of Be in) to 1 My (low sedimentation rates such as found in Fe-Mn crusts). They can be estimated to a maximum age of 10 My (limited by about 7 half-lives of 10Be). They can be determined for the global ocean, at sites where seawater is representative of the global composition, such as the circum-Antarctic area, and for individual basins, where the 10Be/9Be ratio is controlled only by the local denudational input. At sites where advection of water strongly alters the dissolved Be budget, the 10Be/9Be ratio is rather a proxy for ocean circulation. Today this is the case in the South Atlantic and the South Pacific. Finally, the approach can also be applied to large endorheic river basins.

The uncertainty at which relative variations in an ocean basin's paleo-10Be/9Be ratio can be converted into relative estimates in paleo-D depends mainly on the quality of the age model and knowledge of the paleo-10Be flux over this interval (amounting to an error of 3-17%). The accuracy of this conversion further depends on the simultaneous variations in sealevel, in the fraction of Be escaping the coastal trap and the fraction of 9Be released upon primary mineral dissolution. At the current state of knowledge, they add a likely uncertainty of 20 to 30%. The largest improvements in this new method will arise from improvements of past 10Be flux variations as function of magnetic field strength, and from studies that explore the sensitivity of the oceanic 10Be/9Be ratio on boundary exchange of Be as is currently explored for other tracers within the “Geotraces” program (Jeandel et al., 2011).

Acknowledgements. We are grateful to Martin Frank, Marcus Christl, and one anonymous reviewer for their careful and constructive reviews that led to a major refinement of the presentation of the proposed proxy.

Appendix 1. The uncertainty of the modern φdel value

The absolute uncertainty on the global value of φdel is determined by propagating the individual uncertainties of the parameters in equation 5. An uncertainty for the 10 9 -7 ( Be/ Be)oc ratio of ± 0.1 × 10 is used, reflecting the typical difference between 10 Be 10 Be shallow and deep water data (Table 1). We contend that Foc and Friv are associated with an uncertainty of 20% as inferred from the typical differences between the local meteoric 10Be fluxes from Field et al. (2006) and Heikkilä et al. (2008) (see also Willenbring and von Blanckenburg (2010b). The mobile fraction of von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 14

9 9Be 9 Be Be freac + f diss is known ± 0.08 (see section 3.2). As the area of global exhoreic basins Ariv is notoriously difficult to assess, we use as uncertainty the difference between the estimate made here from the database of Milliman and Farnsworth (2011) and that of Peucker-Ehrenbrink (2009), which amounts to 15%. We use an 9 uncertainty on [ Be]parent of ± 0.5 ppm (von Blanckenburg et al., 2012). We estimate the uncertainties associated with denudational fluxes (Milliman and Farnsworth, 2011) to be around 20%. While this analysis emphasises the need to obtaining estimates of φdel with high-fidelity we note that these systematic uncertainties do not 10 9 affect estimates in relative changes of paleo-denudation rate from paleo-( Be/ Be)oc. Instead it is the way these parameters potentially change with D that affect paleo-D estimates. These uncertainties were calculated in section 4.2.

Supplementary Material: Supplement S1: Compilation of seawater dissolved 10Be/9Be ratios Supplement S2: Denudation rate summary Supplement S3: Estimating ocean dust input Supplement S4: Calculating the effect of ocean water advection on a basins 10Be/9Be ratio Supplement S5: An adsorption model for estimating te dependence of 9Be delivery

(φdel) on sediment flux in the coastal zone

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 15

Table 1: Modeled and measured ocean 10Be/9Be ratios

10 10 9 Be ( Be/ Be)oc measured measured F 2 10 9 3 10 9 3 Ocean Basin Aoc Ariv Driv riv,oc model ± Be/ Be ± Be/ Be ± atoms km2 km2 t km-2 y-1 km-2 y-1 Shallow Deep

North Pac >0° 8.30 × 107 1.25 × 107 314 1.20 × 1016 1.23 × 10-7 2.45 × 10-8 1.47 × 10-7 1.6× 10-8 1.19 × 10-7 3.2 × 10-9 South Pac < 0° 9.40 × 107 1.27 × 106 380 1.00 × 1016 9.37× 10-7 1.87 × 10-7 North Atlantic 4.80 × 107 2.49 × 107 144 1.20 × 1016 7.93 × 10-8 1.59 × 10-8 4.64 × 10-8 3.6 × 10-9 6.07 × 10-8 1.6 × 10-9 South Atlantic 4.10 × 107 1.00 × 107 49 8.00 × 1015 3.24 × 10-7 6.48 × 10-8 1.27 × 10-7 2.1 × 10-8 7.36 × 10-8 2.0 × 10-10 Indian1 7.40 × 107 1.41 × 107 406 8.00 × 1015 5.04 × 10-8 1.01 × 10-8 7.20 × 10-8 1.3 × 10-9 7.20 × 10-8 1.3 × 10-9 Arctic 1.70 × 107 1.75 × 107 36 4.00 × 1015 5.45 × 10-8 1.09 × 10-8 4.7 × 10-8 6.2× 10-9 6.6 × 10-8 2.9 × 10-9 Mediterranean 2.50 × 106 3.94 × 106 163 1.00 × 1016 2.05 × 10-8 4.11 × 10-9 1.1 × 10-8 2.4 × 10-9 1.0 × 10-8 2.2 × 10-11 All Oceans4 3.60 × 108 8.43 × 107 184 9.67 × 1015 1.11 × 10-7 2.22 × 10-8 9.71 × 10-8 9.4 × 10-9 8.38 × 10-8 3.0 × 10-9 All Oceans4 without S Pac 2.66 × 108 8.31 × 107 181 9.57 × 1015 8.42 × 10-8 1.68 × 10-8 9.71 × 10-8 9.4 × 10-9 8.38 × 10-8 3.0 × 10-9

Measured 10Be/9Be ocean data (average and standard deviation) are derived from

individual seawater profiles (Supplementary Table S1). Ariv and Driv are calculated from global suspended and dissolved river loads (Milliman and Farnsworth, 2011) as 10 9 given in Supplementary Table S2. Model ( Be/ Be)oc is calculated using a global 9Be 9 Be 10 Be 10 Be freac + f diss of 0.2, a global φdel of 0.063, and the basins specific Foc = Friv as extracted from global flux maps (Willenbring and von Blanckenburg, 2010b).

(1) Indian ocean 10Be/9Be data is from the authigenic phase of surface sediment

samples (Bourlès et al., 1989) as no seawater data is available for this basin.

10 9 (2) Uncertainties on model ( Be/ Be)oc do not contain systematic errors of model parameters to avoid obscuring inter-basin differences; only the 20% uncertainty for

the basins Driv was propagated.

(3) Uncertainties of the average measured 10Be/9Be ratios given are one standard error multiplied with Student’s t-factor at the 60% confidence level.

(4) Measured global ocean 10Be/9Be ratios were calculated on an area-weighted basis (Supplement S1). Not using ocean volume instead adds only a minor uncertainty as only very few water profiles exist for 10Be/9Be, where their low number introduces the largest uncertainty on the global average.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 16

Figures

Figure 1: Cartoon showing the fluxes of 9Be and 10Be. The parameters defined in equation (5) are coloured in the cartoon and in the color version of the equation 9Be 9 Be shown here, with the exception of freac + f diss .

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 17

Figure 2: The mobile fraction of 9Be shown as a function of denudation rate D for the rivers used in the global 10Be/9Be compilation of von Blanckenburg et al. (2012). 10 9 Open symbols: ( Be/ Be)diss measured in the dissolved load of rivers where D is from 10 9 river loads (Milliman and Farnsworth, 2011). Closed and grey symbols: ( Be/ Be)reac measured on the reactive phase of Amazon river and tributary sediment, where D is from in situ-produced cosmogenic nuclides (data sources as in von Blanckenburg et al. (2012)). Closed symbols: Andean tributaries and Amazon main stream. Grey symbols: Rivers draining shield areas of which the significance for a global relationship between D and the mobile fraction of 9Be is unclear. Three fits are shown that describe the relationship between D and the mobile fraction of 9Be. Dashed line: a mean fraction of mobile 9Be that is 0.2 ± 0.08 (this value excludes shield rivers); dashed-dotted line: linear fit through all data, where the mobile fraction of 9Be = D[t km-2 y-1] x -0.00013 + 0.31; dotted curve: power law fit through upper envelope of data (four points denoted by circles), where the mobile fraction of 9Be = 2.6 x D[t km- 2 y-1]-0.32. These fits are used in Figure 6 to evaluate how the oceans 10Be/9Be should change with a change in global denudation.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 18

Figure 3: Measured mean ocean basin dissolved 10Be/9Be ratios against the ratio modeled using equation (5) (Table 1). a) “Shallow water” averages dissolved Be measurements from samples up to 1000 m depth; b) “Deep water” samples from beneath 1000 m. Closed symbols: individual basin results where the model ratio is based solely on the denudational Be input. Open symbols in b): model data comprises the effects of both denudational Be input and deep water advection from an adjacent basin (Figure 4, Supplement S4). Grey star: ocean basin area-weighted 10 9 measurements and model ( Be/ Be)oc for the global oceans without the southern Pacific. For the Indian Ocean no dissolved Be data are available: in this case the ratio from leachates of modern surface sediments (Bourlès et al., 1989) was used instead. All modeled values were calculated using a global river mobile fraction of 9 10 9 Be of 0.2 and φdel of 0.063. Uncertainties of measured Be/ Be ratios are one standard error of means of individual seawater depth profiles. As uncertainties 10 Be 10Be 9 9 of F , F , [ Be]parent, the mobile fraction of Be, and φdel are assumed to be riv metoc mostly systematic, meaning they affect all ocean basins to a similar extent, 10 9 propagating them into individual modelled basins ( Be/ Be)oc would only obscure differences. Hence we only propagate the uncertainty on the individual basins’ denudational flux Driv, estimated to be approximately 20%.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 19

10 9 Figure 4. Sensitivity of oceanic Be/ Be ratios to advection of water masses from an 10 9 adjacent ocean basin. Here, for each ocean basin, a mixture of a ( Be/ Be)oc ratio 10 9 10 9 based only on terrigenous inputs (( Be/ Be)denudation, corresponding to ( Be/ Be)oc in equation (5), values in Table 1) and an input from adjacent ocean basins is calculated following a simple binary mixture equation (see Supplement S4). The 10 9 10 9 resulting ratio ( Be/ Be)mixture depends on the Be/ Be ratio of the water advected 9 and on the flux of the advected Be F( Be)advected, as compared to the flux of the 9 10 9 denudational Be F( Be)denudation. For each ocean basin, the Be/ Be would shift along 9 9 a curve with a changing ratio F( Be)advected/ F( Be)denudation. The present status of each ocean basin is shown by black squares. On the left-hand side of the diagram, where 9 9 10 9 F( Be)advected/F( Be)denudation is < 1, ( Be/ Be)mixture is not strongly modified by advection of water, as the advective Be flux is too low. In these basins, changes in D can be evaluated. On the right-hand side of the diagram, where the curves are steep, 10 9 ( Be/ Be)oc is sensitive to the strength of advection of water, as the advective Be flux overwhelmes the denudational flux. In these basins, changes in the strength of deep 9 10 9 water circulation result in changes of F( Be)advected and ( Be/ Be)oc becomes sensitive to changing deep water circulation. For the apparently flat curves (e.g. the Arctic 10 9 basin) ( Be/ Be)oc set by the basins denudation rate is identical to the ratio of the water mass potentially advected into the basin. Here changes in advection will not modify the 10Be/9Be ratio regardless of the advective strength.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 20

10 Figure 5. Illustration of the uncertainty introduced by past variations in F Be that are caused by variations in the magnetic field strength into estimates in relative changes 10 9 of past D from ( Be/ Be)oc extracted from the authigenic phase of ocean sediment. 10 9 To be used as a proxy for D, paleo-( Be/ Be)oc has to be corrected for variations in 10 F Be . The uncertainty introduced by this correction would be zero if samples were taken in a sufficiently small age interval using an accurate age model, at an age at 10 which F Be is known with perfect accuracy. We estimate the relative error arising from the departure from this ideal case as in practice age models of sedimentary sequences are imperfect (denoted by an absolute “Age offset”), and samples integrate over a given interval (denoted by “Integration time”). Larger errors in age models result in larger uncertainty in paleo-D, while higher integration time results in 10 reduced uncertainty as the variations in F Be tend to cancel out. At high integration times, (100 kyrs), such errors are small, regardless of the error on the age model. The curves denoting this error were generated by first calculating the "true" 10Be production flux when averaging 1000 values that were randomly selected over the whole 10Be production record of Christl et al. (2010) at the prescribed age and over the selected integration time. Then, production fluxes were integrated over the same integration time scale, but around an age that is offset (either towards younger or older ages) from the first one by the prescribed "age offset". The relative difference between the two 10Be production fluxes was then calculated and is shown by the mean (solid curves) and standard deviation (grey area is mean ± standard deviation for the "Age offset = 5 kyrs"-curve) were calculated over this set of 1000 values.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 21

Figure 6: Relationship between D and oceanic 10Be/9Be ratios. All curves were 10Be 10Be 6 -2 -1 9 calculated using Foc = Friv = 10 atoms cm y and [ Be]parent = 2.5 ppm. The "Reference"-model (thick solid line) is based on a constant mobile fraction 9Be 9 Be ( freac + f diss ) = 0.2, a constant φdel = 0.063, and the global ocean and exhoreic drainage areas presented in Table 1. In the "Low seawater level (glacial)"-model, the global ocean area is decreased by 1.6 107 km2, and the exhoreic river basin area is increased by the same area (François et al., 1998). For the "φdel depends on D"- model, φdel was expressed as a function of sediment delivery to the global ocean (Supplement S5), where sediment delivery is assumed to vary proportionally to D. In the coastal zone, an increase in river sediment delivery results in two competing effects: a lowering of the partition coefficient of Be between the reactive and dissolved phases in the coastal ocan (following (Hawley et al., 1986)); and an increase in the availability of suspended sediment in the coastal ocean. For the 9Be 9 Be 9Be 9 Be " freac + f diss depends on D"-model, freac + f diss was assumed to follow the power- 9Be 9 Be law fitting the upper envelop of the freac + f diss -D relationship of large rivers (Figure 2), which is likely an overestimate. The two horizontal lines show a possible 10 9 minimimum and maximum value of today’s ( Be/ Be)oc, corresponding to the typical intra-ocean basin variability observed in seawater (Table 1). The vertical grey bar describes the range of D values that could be inferred from each of these 10 9 ( Be/ Be)oc.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 22

References Amiotte-Suchet, P.A., Probst, J.L., Ludwig, W., 2003. Worldwide distribution of continental rock lithology: Implications for the atmospheric/soil CO2 uptake by continental weathering and alkalinity river transport to the oceans. Global Biogeochemical Cycles 171038, doi:10.1029/2002GB001891. Anderson, R.F., Lao, Y., Broecker, W.S., Trumbore, S.E., Hofmann, H.J., Wölfli, W., 1990. Boundary scavenging in the Pacific Ocean: a comparison of 10Be and 231Pa. Earth and Planetary Science Letters 96, 287-304. Arsouze, T., Dutay, J.C., Lacan, F., Jeandel, C., 2009. Reconstructing the Nd oceanic cycle using a coupled dynamical - biogeochemical model. Biogeosciences 6, 2829-2846. Bourlès, D., Raisbeck, G.M., Yiou, F., 1989. 10Be and 9Be in marine sediments and their potential for dating. Geochimica Cosmochimica Acta 53, 443-452. Brown, E.T., Measures, C.I., Edmond, J.M., Bourles, D.L., Raisbeck, G.M., Yiou, F., 1992. Continental inputs of Beryllium to the oceans. Earth and Planetary Science Letters 114, 101-111. Burton, K.W., Vance, D., 2000. Glacial-interglacial variations in the neodymium isotope composition of seawater in the Bay of Bengal recorded by planktonic foraminifera. Earth and Planetary Science Letters 176, 425-441. Carcaillet, J.T., Bourlès, D.L., Thouveny, N., 2004. Geomagnetic diploe moment and 10Be production rate intercalibration from authigenic 10Be/9Be for the last 1.3 Ma. Geochemistry Geophysics Geosystems doi:10.1029/2003GC000641, 13. Channell, J.E.T., Xuan, C., Hodell, D.A., 2009. Stacking paleointensity and oxygen isotope data for the last 1.5 Myr (PISO-1500). Earth and Planetary Science Letters 283, 14-23. Christl, M., Lippold, J., Steinhilber, F., Bernsdorff, F., Mangini, A., 2010. Reconstruction of global Be-10 production over the past 250 ka from highly accumulating Atlantic drift sediments. Quaternary Science Reviews 29, 2663-2672. Field, C.V., Schmidt, G.A., Koch, D., Salyk, C., 2006. Modeling production and climate-related impacts on 10Be concentration in ice cores. Journal of Geophysical Research 111, doi:10.1029/2005JD006410. François, L.M., Delire, C., Warnant, P., Munhoven, G., 1998. Modelling the glacial– interglacial changes in the continental biosphere. Global and Planetary Change 16– 17, 37-52. Frank, M., Porcelli, D., Andersson, P.S., Baskaran, M., Björk, G., Kubik, P.W., Hattendorf, B., Günther, D., 2009. The dissolved beryllium isotope composition of the Arctic Ocean. Geochimica Cosmochimica Acta 73, 6114-6133. Frank, M., Schwarz, B., Baumann, S., Kubik, P.W., Suter, M., Mangini, A., 1997. A 200 kyr record of cosmogenic radionuclide production rate and geomagnetic field intensity from 10Be in globally stacked deep-sea sediments. Earth and Planetary Science Letters 149, 121-129. Ganachaud, A., Wunsch, C., 2000. Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature 408, 453-457. Gourlan, A.T., Meynadier, L., Allegre, C.J., Tapponnier, P., Birck, J.L., Joron, J.L., 2010. Northern Hemisphere climate control of the Bengali rivers discharge during the past 4 Ma. Quaternary Science Reviews 29, 2484-2498. Grew, E.S., 2002. Beryllium in Metamorphic Environments (emphasis on aluminous von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 23 compositions). Reviews in Mineralogy and Geochemistry 50, 487-549. Hawley, N., Robbins, J.A., Eadie, B.J., 1986. The partitioning of 7beryllium in fresh water. Geochimica Cosmochimica Acta 50, 1127-1131. Heikkilä, U., Beer, J., Feichter, J., 2008. Meridional transport and deposition of atmospheric 10Be. Atmospheric Chemistry and Physics Discussions 8, 16819-16849. Igel, H., von Blanckenburg, F., 1999. Lateral mixing and advection of reactive isotope tracers in ocean basins: numerical modeling. Geochemistry, Geophysics, Geosystems 1, 1999GC000003. Jeandel, C., Peucker-Ehrenbrink, B., Jones, M.T., Pearce, C.R., Oelkers, E.H., Godderis, Y., Lacan, F., Aumont, O., Arsouze, T., 2011. Ocean margins: The missing term in oceanic element budgets? EOS 92, 217-224. Knudsen, M.F., Henderson, G.M., Frank, M., Mac Niocaill, C., Kubik, P.W., 2008. In- phase anomalies in Beryllium-10 production and palaeomagnetic field behaviour during the Iceland Basin geomagnetic excursion. Earth and Planetary Science Letters 265, 588-599. Kusakabe, M., Ku, T.L., Southon, J.R., Measures, C.I., 1990. Beryllium isotopes in the ocean. Geochemical Journal 24, 263-272. Kusakabe, M., Ku, T.L., Southon, J.R., Shao, L., Vogel, J.S., Nelson, D.E., Nakaya, S., Cusimano, G.L., 1991. Be isotopes in rivers/estuaries and their oceanic budgets. Earth and Planetary Science Letters 102, 265-276. Kusakabe, M., Ku, T.L., Southon, J.R., Vogel, J.S., Nelson, D.E., Measures, C.I., Nozaki, Y., 1987. Distribution of 10Be and 9Be in the Pacific Ocean. Earth and Planetary Science Letters 82, 231-240. Lacan, F., Jeandel, C., 2005. Neodymium isotopes as a new tool for quantifying exchange fluxes at the continent-ocean interface. Earth and Planetary Science Letters 232, 245-257. Lao, Y., Anderson, R.F., Broecker, W.S., Trumbore, S.E., Hofmann, H.J., Wölfli, W., 1992. Transport and burial rates of 10Be and 231Pa in the Pacific Ocean during the Holocene period. Earth and Planetary Science Letters 113, 173-189. Li, G., Elderfield, H., 2013. Evolution of carbon cycle over the past 100 million years. Geochim. Cosmochim. Acta 103, 11-25. McDonald, A., Wunsch, C., 1996. An estimate of global ocean circulation and heat fluxes. Nature 382, 436-439. Measures, C.I., Ku, T.L., Luo, S., Southon, J.R., Xu, X., Kusakabe, M., 1996. The distribution of 10Be and 9Be in the South Atlantic. Deep-Sea Research I 43, 987-1009. Milliman, J.D., Farnsworth, K.L., 2011. River discharge to the coastal ocean. A global synthesis. Cambridge University Press, Cambridge. Oelkers, E.H., Jones, M.T., Pearce, C.R., Jeandel, C., Eiriksdottir, E.S., Gislason, S.R., 2012. Riverine particulate material dissolution in seawater and its implications for the global cycles of the elements. Cr Geosci 344, 646-651. Peucker-Ehrenbrink, B., 2009. Land2Sea database of river drainage basin sizes, annual water discharges, and suspended sediment fluxes. Geochemistry Geophysics Geosystems 10. Raiswell, R., 2006. Towards a global highly reactive iron cycle. Journal of Geochemical Exploration 88, 436-439. Rempfer, J., Stocker, T.F., Joos, F., Dutay, J.C., Siddall, M., 2011. Modelling Nd- von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 24 isotopes with a coarse resolution ocean circulation model: Sensitivities to model parameters and source/sink distributions. Geochim. Cosmochim. Acta 75, 5927- 5950. Roy-Barman, M., Coppola, L., Souhaut, M., 2002. Thorium isotopes in the western Mediterranean Sea: an insight into the marine particle dynamics. Earth and Planet. Sc. Lett. 196, 161-174. Valet, J.P., Meynadier, L., Guyodo, Y., 2005. Geomagnetic dipole strength and reversal rate over the past two million years. Nature 435, 802-805. von Blanckenburg, F., Bouchez, J., Wittmann, H., 2012. Earth surface erosion and weathering from the 10Be (meteoric)/9Be ratio. Earth and Planetary Science Letters 351-352, 295-305. von Blanckenburg, F., Igel, H., 1999. Lateral mixing and advection of reactive isotope tracers in ocean basins: observations and mechanisms. Earth and Planetary Science Letters 169, 113-128. von Blanckenburg, F., O'Nions, R.K., 1999. Response of beryllium and radiogenic isotope ratios in Northern Atlantic Deep Water to the onset of Northern Hemisphere Glaciation. Earth and Planetary Science Letters 167, 175-182. von Blanckenburg, F., O'Nions, R.K., Belshaw, N.S., Gibb, A., Hein, J.R., 1996. Global distribution of Beryllium isotopes in deep ocean water as derived from Fe-Mn crusts. Earth and Planetary Science Letters 141, 213-226. Willenbring, J.K., von Blanckenburg, F., 2010a. Long-term stability of global erosion rates and weathering during late-Cenozoic cooling. Nature 465, 211-214. Willenbring, J.K., von Blanckenburg, F., 2010b. Meteoric cosmogenic Beryllium-10 adsorbed to river sediment and soil: applications for Earth-surface dynamics. Earth Science Reviews 98, 105-122. Wilson, D.J., Piotrowski, A.M., Galy, A., McCave, I.N., 2012. A boundary exchange influence on deglacial neodymium isotope records from the deep western Indian Ocean. Earth and Planetary Science Letters 341, 35-47. Xu, X., 1994. Geochemical studies of beryllium isotopes in marine and continental natural systems University of Southern California, LA p. 315. Ziegler, L.B., Constable, C.G., Johnson, C.L., Tauxe, L., 2011. PADM2M: a penalized maximum likelihood model of the 0-2 Ma palaeomagnetic axial dipole moment. Geophysical Journal International 184, 1069-1089.

von Blanckenburg & Bouchez (2014): River fluxes to the sea from the ocean’s 10Be/9Be ratio 25

Online Supplement to: River fluxes to the sea from the ocean’s 10Be/9Be ratio

Friedhelm von Blanckenburg and Julien Bouchez

Supplement S1: Compilation of seawater dissolved 10Be/9Be ratios

Seawater 10Be/9Be ratios were compiled in Table S1 from published depth profiles (Brown et al., 1992; Frank et al., 2008; Ku et al., 1990; Kusakabe et al., 1987; Measures et al., 1996; Xu, 1994). We group our data into “shallow water”, comprising all measurements up to 1000 m depth, essentially thermocline water, and “deep water”, comprising all measurements between 1000 m water depth and the seafloor. No dissolved Be has been measured to date in the Indian Ocean. We instead use the 10Be/9Be ratio measured on leachates of deep ocean surface sediments from the Indian Ocean (Bourlès et al., 1989), assumed to represent the isotopic composition of modern bottom waters. Also no 10Be /9Be ocean data is available yet for the Southern Pacific which is consequently not included in the global value. Measured global ocean 10Be/9Be ratio (Table 1 in main text) were calculated on an area-weighted basis. An accurate estimate of the global ocean 10Be/9Be ratio would ideally be calculated on a Be mass-weighted basis. However, as only very few water profiles exist for Be our approach is rather to calculate the global surface (depth < 1000 m) ocean 10Be/9Be ratios through a simple area-weighting, which adds only a minor inaccuracy. This approach is valid since the main inputs of Be to the global ocean are delivered to the surface ocean, and as Be concentration is fairly homogeneous across all surface ocean basins.

Supplement S2: Denudation rate compilation

10 9 10 9 In order to compare model ( Be/ Be)oc estimated from the source fluxes as defined in equation (5) with Be/ Be ratios published for dissolved Be in seawater, we compiled the denudational flux of solid and dissolved matter discharged into each ocean basin from the global river flux database of Milliman and Farnsworth (2011). To this end, the ocean basins were delineated as defined by a distinct and uniform 10Be/9Be ratio, resulting from mixing of the different riverine inputs in ocean gyres (Igel and von Blanckenburg, 1999; von Blanckenburg and Igel, 1999). We define 6 ocean basins (Table S2) for which data on dissolved 10Be/9Be exist: North Pacific, North Atlantic, South Atlantic, Indian (for which we use chemical extract of seafloor sediment as a proxy for seawater dissolved 10Be/9Be (Bourlès et al., 1989)), Arctic, and Mediterranean Sea.

We first sorted the 1531 rivers of Milliman and Farnsworth (2001)’s database as a function of the ocean region where rivers discharge, using the "Ocean" entry of the database. The 76 "Ocean" entries were then grouped into the 6 ocean basins (Table S2) defined by the 10Be/9Be data. The Equator was considered to be the limit between North and South Pacific. Patterns of oceanic circulation suggest that the Bering Strait supplies water to the North Pacific, hence the Bering Sea was included in the North Pacific. The limit between the North Atlantic and the South Atlantic was set at 5ºS. The latitude of this separation was identified from both the distinct 10Be/9Be ratios and from patterns of oceanic circulation (McDonald and Wunsch, 1996). Therefore, with this definition the Amazon discharges in the North Atlantic, while the Congo discharges in the South Atlantic. Assigning the Congo to the North Atlantic increases the sediment flux it receives by 10% while it decreases that of the South Atlantic by 20%. This shift introduces a negligible difference in the estimate 10 9 of ( Be/ Be)oc. Following Milliman & Farnsworth (2011)’s delineation, the western limit of the South Atlantic is located at ca. 54ºS 68ºW (Grande River in Argentina). The easternmost river discharging to the South Atlantic is the Diep in South Africa (ca. 34ºS 18ºE). The relatively closed water bodies of Southeastern Asia (Gulf of Thailand, Singapore Strait, Karimata Strait, Java Sea, Banda Sea, Timor Sea, Seram Sea, Celebes Sea, Sulu Sea, Makassar Strait, Arafura Strait, and Gulf of Papua) were considered to discharge into the Indian Ocean through the Indonesian throughflow. The Black Sea was considered to be a closed basin and hence was not agglomerated to the Mediterranean Sea.

The river drainage area, available for all of the 1531 rivers, was summed up for each ocean basin (Table S2). Theses areas agree within 15% with those estimated by Peucker-Ehrenbrink et al. (2009), except for the Mediterranean Sea and for the total Atlantic Ocean. The total exhoreic area is 84 millions km3, 15% lower than Peucker-Ehrenbrink et al (2009)’s total area. Sediment and dissolved fluxes were summed over the available data for each ocean basin. The specific sediment (dissolved) yields were calculated for each ocean basin by dividing the sediment (dissolved) flux by the sum of the drainage areas of river basins for which the sediment (dissolved) fluxes are available. The two specific fluxes were added to compute the ocean basin-specific denudation rates. Results are presented in Table S2.

Whether the river flux database is representative can be assessed by calculating the proportion of exhoreic land area for which flux data exists. The percentage of drainage area covered by the discharge data is above 94% for all basins. In all cases sediment flux data is available for more than 70% of the drainage, and dissolved flux data is available for more than 76% of the total drainage area (Table S2).

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 1 -

Supplement S3: Estimating ocean dust input

Beryllium can be transferred from the continents to the ocean through dust. However, von Blanckenburg et al. (1996) showed that the ocean’s dissolved 9Be budget can be balanced by dust as the only contributing source if unlikely high degrees of dust dissolution (>50%) upon entry in seawater are invoked. Willenbring and von Blanckenburg (2010) complemented this analysis by including the substantial amount of meteoric 10Be typically associated with dust, and found that an unrealistic 100% of 10Be release from dust is necessary to account for the oceanic budget of 10Be. We can evaluate the role of dust as significant Be source in the ocean in the light of the present mass balance model. First, meteoric 10Be loading onto dust is accounted for by the river mass balance model, in the way that the continental 10Be 10Be transferred by dust cannot exceed its meteoric depositional flux F , except if a significant flux of dust were sourced metriv in endorheic areas. Second, the annual depositional dust flux to the ocean, as estimated by Jickels et al. (2005), is 0.45 × 9 -1 9 -1 9 9Be 9 Be 10 t y , which compares to 15.4 × 10 t y of river input. Assuming that [ Be]parent and freac + fdiss are identical for the 9 river and the dust sources, then 3% of the Be delivered to the ocean is from dust. It may be argued that φdel is higher for dust than for riverine material, as the latter is delivered directly into the surface open ocean without the need to escape the oceans boundaries. However, as most dust-borne Be will be located in secondary clay minerals that are relatively 9Be 9 Be resistant to dissolution, the dust freac + fdiss is probably much lower than 0.2 in rivers. Altogether, the total dust contribution to the dissolved 9Be in the ocean is most likely lower than a few percent.

Supplement S4: Calculating the effect of ocean water advection on a basins 10Be/9Be ratio

Be budgets in a given ocean basin are controlled not only by terrigenous inputs but also by advection of Be from adjacent ocean basins through deep and shallow ocean circulation. This effect is most profound in those basins in which the 9 9 10 9 advective Be flux dominates over the denudational Be flux. In these basins, paleo-( Be/ Be)oc from the authigenic phase in the sedimentary record will trace changes in paleo-circulation. To assess in what basins this might be the case, we consider that in a given ocean basin, the 10Be/9Be ratio results from a mixture between two end-members :

⎛ 10 Be⎞ (10 Be) + (10 Be) (A1) ⎜ ⎟ = denudation advected ⎝ 9 Be ⎠ 9 Be + 9 Be mixture ()denudation ()advected

Where (10Be) and (9Be) denote inventories. Eq. (A1) can be re-arranged :

⎛ 10 Be⎞ (A2) ⎜ ⎟ ()9 Be ⎝ 9 Be ⎠ 1+ advected ⋅ advection ⎛ 10 Be⎞ ()9 Be ⎛ 10 Be⎞ ⎜ ⎟ denudation ⎜ ⎟ ⎝ 9 Be ⎠ ⎝ 9 Be ⎠ mixture = denudation ⎛ 10 Be⎞ 9 Be ()advected ⎜ 9 ⎟ 1+ Be 9 ⎝ ⎠ denudation Be ()denudation

9 9 Where the ratio ( Be)advected/( Be)denudation is the ratio between the flux of advected Be and the flux of terrigenous Be. In 10 9 case several water masses from different oceans are advected (e.g. for the Southern Atlantic), a "total" ( Be/ Be)advected is calculated as the weighted average of the inputs, following equation (A2).

The parameters used for the calculations are presented in Table S3. Advective fluxes are from Ganachaud and Wunsch (2000), except for the influx into the Arctic ocean, which is from Beszczynska et al. (2011). The 10Be/9Be ratios are those measured as presented in Table 1. Only for the Antarctic and South Pacific oceans, for which no disoolved Be data exist, were the values estimated to be 1× 10-7. 9Be seawater concentrations were taken from various sources (Brown et al., 9 1992; Frank et al., 2008; Ku et al., 1990; Kusakabe et al., 1987; Measures et al., 1996; Xu, 1994 ). F( Be)denudation was calculated using denudation rates and continental areas from from Milliman and Farnsworth’s (2011) database (as 9 9Be 9 Be reported in Table 1), a [ Be]parent concentration of 2.5 ppm, a freac + fdiss of 0.2 and a φdel of 0.063.

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 2 -

9 Supplement S5: An adsorption model for estimating the dependence of Be delivery to the ocean φdel on sediment flux into the coastal zone

In order to close the global ocean 10Be/9Be budget, the fraction of terrigenous Be (10Be and 9Be) delivered to the open ocean as dissolved flux relative to the sum of dissolved plus reactive flux (adsorbed onto particles or present in secondary precipitates), called here φdel, has to be equal to 0.063. The rest of the riverine Be is thought to be trapped in the coastal zone of the ocean, consisting mostly of estuarine and shelf areas. In this appendix we develop a more mechanistic view that accounts for the observed partial transfer of Be through the coastal ocean. Our aim is (1) to establish a relationship between φdel and measurable properties such as concentrations or fluxes; with this relationship we will then perform a first-order test of the inferred φdel value, and (2) derive a link between the value of φdel and continental denudation D. This latter aim is especially important in order to evaluate how φdel might have varied in the past with varying sediment discharge to the ocean, a critical aspect for the validity of oceanic 10Be/9Be ratios as a proxy for past rates of terrigenous inputs.

Terrigenous Be is exported into the open ocean (where it mixes with meteoric 10Be directly deposited onto the ocean) in the dissolved form only, while reactive Be sinks with settling particles to coastal sediments. The critical process governing the flux of Be export to the open ocean is therefore its partitioning between the dissolved and reactive phases in the coastal ocean. Assuming that this partitioning is governed predominantly by adsorption, and to a lesser extent by more complex processes such as co-precipitation, Be exchanges between the reactive and the dissolved phase and adjusts according to the partition coefficient in the coastal ocean, Be -1 Kd (coast) (in L kg ), defined as:

Be [Be]reac (coast) (A3) Kd (coast) = [Be]diss(coast)

-1 where [Be]reac(coast) and [Be]diss(coast) are the concentration of Be in the reactive phase (e.g. in mg kg ) and in -1 Be water (e.g. in mg L ) in the coastal ocean. As isotope fractionation can be neglected, Kd (coast) has the same Be value for both Be isotopes. We define now γ diss(coast) as the fraction of dissolved Be compared to dissolved plus reactive Be. It can be shown that:

(A4) Be 1 γ diss = Be 1+ Kd (coast)⋅ P(coast) where P(coast) is the suspended particle concentration in coastal waters.

Be Next we explore how γ diss(coast) is linked to φdel. As φdel is a ratio of fluxes, we design a simple steady-state box Be model for the coastal ocean. Rivers deliver reactive and dissolved Be to the coastal ocean at a flux Freac+diss(riv) (in kg y-1). After equilibration under conditions prevailing in the coastal ocean, only dissolved Be is exported to Be Be the open ocean through water advection at a flux Fdiss(oc), while reactive Be is exported at a flux Freac (sed) to coastal sediments. These fluxes can be considered as fluxes of 10Be, 9Be, or elemental Be fluxes, since no fractionation occurs during Be exchange, advection, or sedimentation. The steady-state mass-balance equation for the coastal ocean box is:

Be Be Be Freac+diss(riv) = Freac (sed) + Fdiss(oc) (A5)

Be Be We further assume that both Fdiss(oc) and Freac (sed) follow first-order kinetic laws, with Be Be Be Be Be Be Fdiss(oc) = kadv ⋅ Mdiss(coast) and Freac (sed) = ksed ⋅ Mreac (coast), where Mdiss(coast) and Mreac (coast) are the -1 inventories (in kg) of reactive and dissolved Be in the coastal ocean, respectively. kadv (in y ) is related to water residence time in the coastal ocean through kadv = 1/τwat, and hence quantifies how fast water is transported to the open ocean. ksed is related to the residence time of suspended sediments in the coastal ocean (kadv = 1/τsed). With Be Be Be the total mass of Be in the coastal ocean Mreac+diss(coast) = Mreac (coast) + Mdiss(coast), and re-arranging, equation (A5) can be rewritten:

Be Be Freac+diss(riv) Mdiss(coast) (A6) Be = ksed + ()kadv − ksed ⋅ Be Mreac+diss(coast) Mreac+diss(coast)

With the terminology of the box model developed here, φdel can be expressed in terms of eq. (A6) as:

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 3 -

F Be oc k ⋅ M Be coast M Be coast M Be coast (A7) diss( ) adv diss( ) diss( ) reac +diss( ) φdel = Be = Be = kadv ⋅ Be ⋅ Be Freac +diss(riv) Freac+diss(riv) Mreac +diss(coast) Freac+diss(riv)

Be M coast Be The ratio diss( ) in equations (A6) and (A7) is actually equal to γ (coast), by definition. Therefore, Be diss Mreac+diss(coast) combining equations (A6) and (A7), and re-arranging:

1 φ = (A8) del ⎛ ⎞ ksed 1 1+ ⋅⎜ Be −1⎟ kadv ⎝ γ diss ⎠

Be Be Eq. (A8) relates φdel and γ diss(coast). We can now use equation (A4) to link φdel with Kd (coast), which is a measurable property of Be behaviour in seawater:

1 (A9) φdel = ksed Be 1+ ⋅ Kd (coast)⋅ P(coast) kadv

Be For a given value of Kd (coast), equation (A9) prescribes that at high values of P(coast) Be will be preferentially partitioned onto the solids, and φdel should take low values as less Be is readily exported in the Be dissolved form to the open ocean. However, Kd (coast) is not uniform. For Be as for many reactive metals, its value varies with suspended particle concentration (Honeyman and Santschi, 1992) and is lower in seawater than in freshwater (Hawley et al., 1986). This so-called particle concentration effect describes a decline in partition coefficients (Kd) as the concentration of suspended particulate matter increases. This anomaly has been attributed to a variety of causes, but most often to the existence of colloidal forms of the adsorbate, which are included in the dissolved fraction when calculating Kd (Benoit and Rozan, 1999). Hence at high value of P(coast) the value Be of Kd (coast) decreases. Hence variations in particle concentration in the coastal ocean, that might result from varying D and thus supply of river sediment, results in two competing effects that might partially compensate each other. In order to assess the net effect of changes in D on φdel, we first seek for a relationship between P(coast) and the concentration of sediment in rivers, P(riv). Assuming that the inventory of suspended particles in the coastal ocean Msed(coast) is at steady-state, the flux of sediment to the coastal ocean Fsed(riv) is:

Fsed (riv) = k sed ⋅ M sed (coast) (A10)

Similarly, the steady-state masse balance for water in the coastal ocean results in:

Fwat (riv) = kwat ⋅ M wat (coast) (A11) with Fwat(riv) the water river discharge to the coastal ocean and Mwat(coast) the volume of the coastal ocean. Ratioing equations (A10) and (A11), the concentrations of sediment in riverine water and coastal water is:

k τ (A12) P(riv) = sed ⋅ P(coast) = wat ⋅ P(coast) kadv τ sed

Eq. (A12) shows that as water resides longer in the coastal zone than sediment (i.e. if the coastal zone is a compartment of rapid sedimentation), P(riv) is higher than P(coast). In other words, river sediments are diluted upon entry in the coastal ocean.

Finally, eqs. (A9) and (A12) can be combined to yield:

1 (A13) φdel = Be 1+ Kdiss(coast)⋅ P(riv)

We now first use eq. (A13) to perform a first-order check of (a) our conceptual coastal ocean model, and (b) our Be -4 -1 inferred global φdel value. Solving eq. (A13) for Kd (coast) with φdel = 0.063 and P(riv) = 5.10 kg L Be 4 -1 (determined from Milliman and Farnsworth’s (2011) database), Kd (coast) is around 2.4 × 10 L kg . Taking Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 4 - the “concentration effect” on Kd into account, we follow the empirical relationship reported by Hawley et al. Be -1 (1986) between Kd (coast) and particle concentration in seawater P(coast) in mg L : Be 5 −0.805 K d (coast) =1.85×10 P(coast) (A14) this would correspond to a P(coast) value of ca. 10 mg L-1. This concentration is of a reasonable order for the coastal ocean (estuaries, shelves). We conclude that our value φdel of 0.063 is in a reasonable range.

Second, we use eq. (A13) to assess the sensitivity of the ocean 10Be/9Be ratio to changes in denudation rate in the case where φdel is not constant but affected by changing sediment delivery to the coastal ocean. To to so, we Be assume that P(riv) and P(coast) vary linearly with D, and that Kd (coast)varies with P(coast) following the empirical relationship reported by Hawley et al. (1986). Therefore φdel can be calculated as a function of D. The result of this calculation is shown in Fig. 6 of the main text.

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 5 -

Table S1: Compilation of Seawater dissolved 10Be/9Be ratios

Maximum 10Be/9Be Standard 10Be/9Be Standard Author Station Latitude Longitude depth (m) Shallow deviation/ Deep deviation/ 60% error 60% error

Pacific 1.47E-07 4.26E-08 1.19E-07 2.06E-08 1.60E-08 3.16E-09 Xu 1994 Station F 52.60N 144.56W 4110 1.24E-07 2.65E-08 1.16E-07 8.37E-09 Xu 1994 StationB 49.40N 162E 5322 1.49E-07 7.39E-08 1.02E-07 4.99E-09 Xu 1994 Station C 53.31N 177.38E 3749 1.14E-07 3.24E-08 1.04E-07 8.81E-09 Kusakabe 1987 Station 5 25.00N 169.59E 6013 1.93E-07 7.60E-08 1.22E-07 1.16E-08 Kusakabe 1987 St 11 17.00N 117.58W 3950 1.00E-07 1.15E-08 1.09E-07 1.27E-08 Kusakabe 1987 St21 2.00N 117W 3950 2.03E-07 3.05E-08 1.58E-07 2.52E-08

Atlantic South of 20°S 1.27E-07 2.10E-08 7.36E-08 2.02E-10 2.05E-08 1.97E-10 Measures 1996 Station 114 24°40' S 38°21' W 3830 1.42E-07 2.21E-08 7.34E-08 1.53E-08 Measures 1996 Station 141 24°55' S 01°00' E 4152 1.13E-07 3.08E-08 7.37E-08 1.97E-08

Atlantic N of 20°S 4.64E-08 8.51E-09 6.07E-08 3.85E-09 3.58E-09 1.62E-09 Measures 1996 Station 158 01°59' S 04°02' W 4850 5.80E-08 9.06E-09 6.53E-08 6.88E-09 Measures 1996 147 - 158 3.84E-08 2.61E-09 Ku 1990 WBE8 41°32.0'N 63°37.0'W 3411 5.29E-08 1.34E-08 6.15E-08 2.92E-09 Ku 1990 WBE10 41°16.9'N 63°32.0'W 3867 5.78E-08 8.70E-09 Ku 1990 WBE22XC 34°1.0'N 63°0.0'W 5178 4.06E-08 6.31E-09 5.59E-08 2.27E-09 Ku 1990 WBE32 32°2.7'N 75°2.2'W 4123 6.32E-08 7.92E-09 Ku 1990 WBE1-34 33°-42°N 62°-75°W surface 4.24E-08 1.23E-08

Mediterranean 1.06E-08 2.42E-09 1.01E-08 2.22E-11 2.36E-09 2.16E-11 Brown 1992 EastMed 35°29.3'N 14.8°E 3770 1.23E-08 2.45E-09 1.01E-08 2.37E-09 Brown 1992 WestMed 38°0.3'N 8°59.9'E 2700 8.89E-09 1.64E-09 1.02E-08 5.03E-10

Arctic 4.72E-08 3.12E-08 6.64E-08 7.77E-09 6.16E-09 2.92E-09 Frank 2009 1: Fram Strait, Station #26, 1379 m 80°20,28'N, 07°20,73'W 1379 2.81E-08 9.61E-10 6.37E-08 9.10E-09 Frank 2009 2: Leg1, Station #2, 158 m 80°25,85'N, 15°31,05'E 158 6.13E-08 1.81E-08 Frank 2009 3: LegII, Station #1, 497 m 81°16,70'N, 26°22,97'E 497 7.83E-08 2.03E-08

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 6 -

Frank 2009 4: LegII, Station #10, 4025 m 83°46,93'N, 31°57,10'E 4025 9.40E-08 3.47E-08 Frank 2009 4: LegII, Station #11, 4039 m 84°16,87'N, 33°39,81'E 4039 5.42E-08 1.70E-08 8.10E-08 1.14E-08 Frank 2009 LegII, Station #12, 4027 m 84°44,04'N, 35°14,89'E 4027 7.12E-08 1.58E-08 Frank 2009 LegII, Station #20, 4410 m 88°16,70'N, 82°54,31'E 4410 2.86E-08 3.40E-09 Frank 2009 5: LegII, Station #21, 4400 m 88°24,48'N, 95°22,78'E 4400 5.33E-08 2.61E-08 6.92E-08 2.72E-08 Frank 2009 5: LegII, Station #22, 4372 m 88°26,05'N, 109°50,65'E 4372 2.36E-08 1.58E-08 Frank 2009 LegII, Station #24, 2779 m 88°21,85'N, 126°29,76'E 2779 2.08E-08 2.50E-09 Frank 2009 LegII, Station #26, 1472 m 88°08,43'N, 132°33,24'E 1472 2.05E-08 2.40E-09 Frank 2009 LegII, Station #26, 1513 m 87°08,43'N, 132°33,24'E 1513 1.76E-08 2.20E-09 Frank 2009 LegII, North Pole, 4260 m 4260 1.99E-08 3.30E-09 Frank 2009 6: LegII, Station #30, 3985 m 87°54,97'N, 154°22,50'E 3985 4.15E-08 2.26E-08 6.03E-08 7.00E-09 Frank 2009 Station AWS#1, 50 m 71°40,0'N, 154°46,0'W 50 5.30E-09 5.66E-10 Frank 2009 Station AWS#2, 170 m 75°12,5'N, 149°57,0'W 170 2.87E-08 2.50E-08 Frank 2009 Station AWS#3, 3850 m 75°12,5'N, 149°57,0'W 3850 6.28E-08 4.30E-08 6.15E-08 6.34E-09 Frank 2009 Station AWS#4, 3894 m 73°49,5'N, 152°00,7'W 3894 5.85E-08 3.57E-08 6.28E-08 1.07E-08 Frank 2009 Station AWS#5, 1200 m 72°14,7'N, 155°04,5'W 1200 1.29E-07 3.27E-08

Indian 7.20E-08 1.31E-08 1.29E-09 Bourles 1989 MD77-217 11°56.3'5 83°00'E 4 930 8.42E-08 7.60E-09 Bourles 1989 MD76-104 35°50.6'S 58°27.6'E 4 860 7.36E-08 6.50E-09 Bourles 1989 MD76-104 49°49.3'S 51°19.1'E 3 262 5.81E-08 5.20E-09

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 7 -

Table S2: Denudation Rate Summary

Area with water Area with sed. Area with diss. Drainage Water Sed. Diss. Sed. Diss. disch. flux flux. area disch. flux flux yield yield data available data available data available

3 t t % % % Contributing "Oceans" [sensu Milliman & 3 2 km Mt Mt -2 -2 3 2 3 2 3 2 Ocean basin 10 km -1 -1 -1 km km 10 km total 10 km total 10 km total Farnsworth (2011)] y y y -1 -1 y y area area area Pacific > 0º, South China Sea, Taiwan Strait, East China Sea, Tsuchima Strait, Yellow Sea, Sea of North Pacific Japan, Sea of Okhotsk, Bering Sea, Gulf of Alaska, 12529 4802 2997 550 263 51 12232 98 11371 91 10862 87 Puget Sound, San Francisco Bay, Monterrey Bay, Gulf of California, Panama Bay Atlantic > 5ºS, Carribean, Gulf of Mexico, Chesapeake Bay, Gulf of Maine, Gulf of St. Lawrence, Labradot North Atlantic Sea, Norwegian Sea, North Sea, Gulf of Finland, Gulf 24945 11403 2393 783 106 38 24002 96 22123 89 20643 83 of Bothnia, Baltic Sea, English Channel, Irish Sea, Bay of Biscay, Gulf of Guinea South Atlantic Atlantic < 5ºS 10048 1564 292 32 17 3 9820 98 9091 90 8838 88 Mediterranean Mediterranean, Gulf du Lion, Ligurian Sea, Sicilian 3935 371 535 142 22 36 3835 97 3781 96 3460 88 Sea Channel, Ionian Sea, Gulf of Tronto, Sea of Marmara Indian, Red Sea, Arabian Sea, Strait of Ormuz, Persian Gulf, Bay of Bengal, Andama Sea, Singapore Strait, Gulf of Thailand, Karimata Strait, Java Sea, Indian Banda Sea, Timor Sea, Seram Sea, Celebes Sea, 11725 3109 2655 349 56 30 11030 94 8238 70 8901 76 Sulu Sea, Makassar Strait, Arafura Sea, Gulf of Papua, Mallaca Strait, Gulf of Carpentaria, Tasman Sea, Foveaux Strait, Great Australian Bight Arctic, Barent Sea, Kara Sea, Laptev Sea, Siberian Arctic Sea, White Sea, Chukchi Sea, Beaufort Sea, Hudson 17526 3603 226 16 20 1 17241 98 13927 79 15958 91 Bay, James Bay, Ungava Bay

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 8 -

Table S3: Influence of water advection through thermohaline circulation on the oceanic 10Be/9Be ratios

10 9 9 10 9 Total 9 9 10 9 ( Be/ Be) Advection Advected [ Be] ( Be/ Be) 10 9 F( Be)advected F( Be)denudation ( Be/ Be)mixture/ Ocean ( Be/ Be) -1 -1 10 9 denudation flux (Sv) from (pM) advected (t y ) (t y ) ( Be/ Be)denudation advected

N. Atl. 8.31E-08 16 S. Atl. (surface) 15 1.27E-07 1.27E-07 68 117 1.20

S. Atl. 3.40E-07 17 N. Atl. (bottom) 25 6.07E-08 8.29E-08 277 16 0.29 22 Antarctic 25 1.00E-07 S. Pac. 9.84E-07 26 Antarctic 25 1.00E-07 1.01E-07 192 16 0.17 1 N. Pac. (bottom) 25 1.19E-07 N. Pac. 1.29E-07 0.5 S. Pac. (top) 15 1.00E-07 1.00E-07 13 128 0.98 1.5 S. Pac. (bottom) 25 1.00E-07 Indian 5.29E-08 11 Antarctic 25 1.00E-07 1.22E-07 146 186 1.57 16 N. Pac. (surface) 15 1.47E-07 Indian* 5.29E-08 11 Antarctic 25 1.00E-07 1.00E-07 78 0.46 1.26 Med 2.15E-08 0.8 N. Atl. (surface) 25 4.64E-08 4.64E-08 6 21 1.25 Arctic 5.71E-7 0.8 Bering Strait 15 1.20E-07 5.91E-08 26 20 2.02 2 Bering Sea Opening 15 5.00E-08 1.8 Fram Strait 28 5.00E-08

*without Indonesian throughflow

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 9 -

References

Benoit, G., Rozan, T.F., 1999. The influence of size distribution on the particle concentration effect and trace metal partitioning in rivers. Geochim. Cosmochim. Acta 63, 113-127. Beszczynska-Moller, A., Woodgate, R.A., Lee, C., Melling, H., Karcher, M., 2011. A Synthesis of Exchanges Through the Main Oceanic Gateways to the Arctic Ocean. Oceanography 24, 82-99. Bourlès, D., Raisbeck, G.M., Yiou, F., 1989. 10Be and 9Be in marine sediments and their potential for dating. Geochimica Cosmochimica Acta 53, 443-452. Brown, E.T., Measures, C.I., Edmond, J.M., Bourles, D.L., Raisbeck, G.M., Yiou, F., 1992. Continental inputs of Beryllium to the oceans. Earth and Planetary Science Letters 114, 101-111. Frank, M., Backman, J., Jakobsson, M., Moran, K., O’Regan, M., King, J., A. Haley, B.A., Kubik, P.W., Garbe- Schönberg, D., 2008. Beryllium isotopes in central Arctic Ocean sediments over the past 12.3 million years: Stratigraphic and paleoclimatic implications. Paleoceanography 23, PA1S02, doi:10.1029/2007PA001478. Ganachaud, A., Wunsch, C., 2000. Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature 408, 453-457. Hawley, N., Robbins, J.A., Eadie, B.J., 1986. The partitioning of 7beryllium in fresh water. Geochimica Cosmochimica Acta 50, 1127-1131. Honeyman, B.D., Santschi, P.H., 1992. The role of particles and colloids in the transport of radionuclides and trace metals in the oceans, in: Buffle, J., van Leeuwen, H.P. (Eds.), Environmental Particles. Environmental Analytical and Physical Chemistry Series. Lewis Publishers, Boca Raton, pp. 379-423. Igel, H., von Blanckenburg, F., 1999. Lateral mixing and advection of reactive isotope tracers in ocean basins: numerical modeling. Geochemistry, Geophysics, Geosystems 1, 1999GC000003. Jickells, T.D., An, Z.S., Andersen, K.K., Baker, A.R., Bergametti, G., Brooks, N., Cao, J.J., Boyd, P.W., Duce, R.A., Hunter, K.A., Kawahata, H., Kubilay, N., laRoche, J., Liss, P.S., Mahowald, N., Prospero, J.M., Ridgwell, A.J., Tegen, I., Torres, R., 2005. Global iron connections between desert dust, ocean biogeochemistry, and climate. Science 308, 67-71. Ku, T.L., Kusakabe, M., Measures, C.I., Southon, J.R., Cusimano, G., Vogel, J.S., Nelson, D.E., Nakaya, S., 1990. Beryllium isotope distribution in the western North Atlantic: a comparison to the Pacific. Deep-Sea Research II 37, 795-808. Kusakabe, M., Ku, T.L., Southon, J.R., Vogel, J.S., Nelson, D.E., Measures, C.I., Nozaki, Y., 1987. Distribution of 10Be and 9Be in the Pacific Ocean. Earth and Planetary Science Letters 82, 231-240. McDonald, A., Wunsch, C., 1996. An estimate of global ocean circulation and heat fluxes. Nature 382, 436-439. Measures, C.I., Ku, T.L., Luo, S., Southon, J.R., Xu, X., Kusakabe, M., 1996. The distribution of 10Be and 9Be in the South Atlantic. Deep-Sea Research I 43, 987-1009. Milliman, J.D., Farnsworth, K.L., 2011. River discharge to the coastal ocean. A global synthesis. Cambridge University Press, Cambridge. von Blanckenburg, F., Igel, H., 1999. Lateral mixing and advection of reactive isotope tracers in ocean basins: observations and mechanisms. Earth and Planetary Science Letters 169, 113-128. von Blanckenburg, F., O'Nions, R.K., Belshaw, N.S., Gibb, A., Hein, J.R., 1996. Global distribution of Beryllium isotopes in deep ocean water as derived from Fe-Mn crusts. Earth and Planetary Science Letters 141, 213-226. Willenbring, J.K., von Blanckenburg, F., 2010. Long-term stability of global erosion rates and weathering during late-Cenozoic cooling. Nature 465, 211-214. Xu, X., 1994. Geochemical studies of beryllium isotopes in marine and continental natural systems University of Southern California, LA p. 315.

Supplement to von Blanckenburg and Bouchez (2013): River fluxes to the sea from the ocean’s 10Be/9Be ratio - 10 -