Clay mineralogy, strontium and neodymium isotope ratios in the sediments of two High Arctic catchments (Svalbard)

The identification of sediment sources to the ocean is a pre-requisite to using ocean sediment cores to extract information on past climate and ocean circulation. Sr and Nd isotopes are classical tools with which to trace source provenance. Yet, despite considerable interest in the Arctic Ocean, the circum-Arctic source regions are poorly characterised in terms of their Sr and Nd isotopic compositions. In this study we present Sr and Nd isotope data from the Paleogene Central Basin sediments of Svalbard, including the first published data of river sediments from Svalbard. 5 The bulk sediments exhibit considerable isotopic variation (εNd0 = -24.2 to -11.9; Sr/Sr = 0.72449 to 0.75243) which can be related to the depositional history of the sediments. In combination with analysis of the clay mineralogy of the sediments, we suggest that the Central Basin sediments were derived from two ‘proto’ sediment sources. One source is Proterozoic sediments derived from Greenland which are rich in illite and have high Sr/Sr and low εNd0 values. The second source is Carboniferous to Jurassic sediments derived from Siberia which are rich in smectite and have low Sr/Sr and high εNd0 10 values. Due to a change in deposition conditions throughout the Paleogene (from deep-sea to continental) the relative proportions of these two sources varies in the Central Basin formations. The modern river suspended sediment isotopic composition is then controlled by modern processes, in particular glaciation, which determines the present-day exposure of the formations and therefore the relative contribution of each formation to the suspended sediment load. This study demonstrates that the Sr and Nd isotopic composition of river sediment from the continents exhibits significant seasonal variation, which almost 15 certainly mirrors longer-term hydrological changes, with implications for source provenance studies based on fixed sources through time.

mid-Eocene with the erosion of the uplifted West Spitsbergen Fold and Thrust Belt whose formation is linked to rifting of the North Atlantic and the separation of Svalbard from Greenland.
The oldest formations exposed in the studied catchments are the Grumantbyen and Hollendardalen Formations, comprising shallow marine sandstones. The Grumantbyen sediments are part of a regressive trend with sediment derived from the east and possibly the north (Helland-Hansen, 1990). The Hollendardalen sandstones were deposited in the west of the basin from the 5 initial uplift of the fold belt and are coeval with the sediments of the lower Frysajodden Formation.
The youngest three formations comprise a regressive sequence with (from oldest to youngest) the Frysjaodden formation comprising fine-grained shales deposited offshore in an open basin; Battfjellet Formation comprising shallow marine sandstone; and Aspelintoppen comprising continental deposits (Helland-Hansen, 1990;Müller and Spielhagen, 1990). The mountain belt is thought to have eroded rapidly (Cui et al., 2011) based on the immaturity of the sandstones (Helland-Hansen, 1990). 10 Detection of pre-Caledonian metamorphic detritus indicates that the mountain belt was eroded down into the basement rocks (Helland-Hansen, 1990). The PETM boundary is near the base of the Frysjaodden Formation (Charles et al., 2011).

Methods
A selection of 18 representative rock and sediment samples were sampled from both catchments. The rock samples were first crushed (jaw crusher) and were subsequently ground to fine powders (rotary disc mill and planetary ball mill). For the 15 sediment samples, only the latter step was required. Suspended sediment (>0.2 µm) was also retrieved from nylon filter papers during water sample collection by washing the filter paper with deionized water and then freeze drying the sample. Part of each suspended sediment sample was treated with 5% HCl to remove carbonates. The leachates were not retained. A sub-set of samples were selected for element and isotopic analysis and they are further described here. For all other samples a brief description is included in Table 2. The central part of the unglaciated catchment is dominated by small angular pieces of shale glycerol-treated, 450 • C heated and 550 • C heated samples. The glycerol-treated and 450 • C spectra were used to obtain semiquantitative clay mineral abundances using the method outlined in Griffin (1971) which uses the peak heights of kaolinite 001 and illite 001 in the 450 • C spectrum and the peak heights of chlorite 004, kaolinite 002, kaolinite 001 and illite 001 in the glycerol treated spectrum to give the relative abundances of kaolinite, chlorite, illite (mica) and expandable layer clay minerals (e.g. smectite and mixed-layer minerals containing smectite).

Chemical and isotopic composition
A selection of solid samples, separated clay fractions and suspended sediments were analysed for the major and trace element chemistry using the following method: approximately 100 mg of material was ashed at 950 • C for 120 minutes. The sample was then digested in a mixture of concentrated hydrofluoric and nitric acids and repeatedly dried down and re-dissolved in 6M HCl. In the final step, the dried down sample was re-dissolved in 2% HNO 3 . Major and trace element concentrations were 10 measured at the University of Cambridge by inductively-coupled plasma optical emission spectrometry (ICP-OES, Agilent Technologies 5100) and quadrupole inductively-coupled plasma mass spectrometry (Q-ICP-MS, Perkin Elmer 63 Nexion 350D), respectively.
Neodymium was separated from the matrix using the method described in (Piotrowski et al., 2009). This method uses two columns. The first column contained Eichrom TRUspec resin which separates out REE from the matrix and the second 15 column contains Eichrom LNSpec resin to isolate Nd. The radiogenic neodymium isotopic composition was measured on a Nu plasma (Nu Instruments, University of Cambridge) multi-collector inductively coupled plasma mass spectrometer (MC-ICP-MS). Samples were run at 50-75 ppb with an APEX ACM sample introduction system. Samples were run in triplicate (three measurements on different days) with each measurement comprising 30 cycles with 10 s integration. Samarium interferences were monitored by measuring mass 149. No interferences were detected and oxides were monitored during tuning to ensure 20 they were well below 1% of the beam size. The exponential law was applied to correct for instrument mass fractionation and all 143 Nd/ 144 Nd ratios were normalised to 146 Nd/ 144 Nd = 0.7219. Standard-sample were bracketing was employed in order to correct for the minor offset with the accepted JNdi-1 value: 0.512060±0.000024 (2SD, n=119) compared with the accepted value of 0.512115 (Tanaka et al., 2000). The USGS shale standard SCo-1 was measured and the 143 Nd/ 144 Nd value of 0.512086±0.000029 (n=3, 2SD) is in agreement with a previously published value of 0.512117±0. 000010 (2σ, n=20 Krogstad 25 et al., 2004). In this study, neodymium ratios are reported as deviations relative to the chondritic uniform reservoir (CHUR, 143 Nd/ 144 Nd = 0.512638, Jacobsen and Wasserburg, 1980). Strontium was separated from the matrix using Biorad mirco bio-spin columns with Eichrom SrSpec resin (Hindshaw, 2011).
The radiogenic strontium isotopic compositions were measured on a Neptune MC-ICP-MS (Thermo, University of Cambridge) and were run at 50 ppb using an APEX sample introduction system. Samples were run in triplicate (three measurements on 30 different days) with each measurement comprising 30 cycles with 8 s integration. 85 Rb was monitored to correct for rubidium interferences on 87 Sr and data were additionally corrected for Kr interferences by measuring 83 Kr. The exponential law was applied to correct for instrument mass fractionation and all 87 Sr/ 86 Sr ratios were normalised to 86 Sr/ 88 Sr = 0.1194. Measure-5 Earth Surf. Dynam. Discuss., https://doi.org/10.5194/esurf-2017-55 Manuscript under review for journal Earth Surf. Dynam. Discussion started: 19 September 2017 c Author(s) 2017. CC BY 4.0 License. ments of NBS 987 gave a 87 Sr/ 86 Sr value of 0.710249±29 (n=27) and the seawater value was 0.709188±24 (n=9), which is within error of the accepted value of 0.709179±8 (Mokadem et al., 2015).

Results
The major and trace element concentrations of the solid samples are provided in Table 1. The measured values are typical for shales (Taylor and McLennan, 1985) and the rare earth element (REE) chondrite normalised element profile of these samples 5 closely follows that of the Post Archaean Australian Shale (PAAS, Table A2). The major element chemistry is very similar to that observed in core samples drilled through the same formations (Fig. 2, Schlegel et al., 2013). Strontium and neodymium concentrations varied from 78 to 139 and 25 to 49 ppm, respectively.
There was a large range in both the strontium and neodymium isotopic compositions of the bulk rocks and sediments: 87 Sr/ 86 Sr = 0.72449 to 0.75243 and εNd 0 = -24.2 to -11.9. In general, samples collected from Fardalen e.g. R01 and G have 10 higher 87 Sr/ 86 Sr and lower εNd 0 values than those samples collected in Dryadbreen e.g. D and O (Table 1).
The clay-sized fraction forms a parallel array to the bulk rock samples in εNd 0 -87 Sr/ 86 Sr space (Fig. 3), with the claysized samples having higher εNd 0 and lower 87 Sr/ 86 Sr values (except for R01). The εNd 0 values of clay fractions were 2.1 to 3.2 epsilon units higher than the corresponding bulk sample and 87 Sr/ 86 Sr values were 1030 to 2030 ppm lower in the clay compared to the bulk, apart from sample R01 where the clay was 1100 ppm higher in the clay than in the bulk. Rubidium, 15 strontium, neodymium and samarium concentrations are comparable in bulk and clay samples (Table 1) suggesting that clays are the main host of these elements. It has been observed that in a compilation of river sediments from all over the world that εNd 0 in the clay fraction is greater than in the silt-sized fraction by an average of 0.8 epsilon units (Bayon et al., 2015).
Fine sediments (as measured by Al/Si ratio) from the Mackenzie River have also been observed to have higher values than coarser sediments (Vonk et al., 2015). The offset in εNd 0 between fine and coarse fractions has been interpreted to reflect the 20 preferential transport of basalt and volcanics in the fine fraction (McLennan et al., 1989;Garçon and Chauvel, 2014;Bayon et al., 2015). A volcanic signal is typically only observed in the first sedimentary cycle, due to the rapid chemical weathering of volcanic particles (McLennan et al., 1989) and therefore, if volcanics are present, they must have been deposited at the same time as the Central Basin sediments. Potential volcanic sources for this period could be the volcanic provinces of North Greenland and Ellesmere Island (58-61 Ma, Jones et al., 2016).

Semi-quantitative clay abundance
Illite, chlorite and kaolinite were present in all the samples analysed. In addition the presence of an expandable layer clay mineral is also evident in the collapse of the XRD signal around 12.7 Å (8 • 2θ Co radiation) between the air-dried and 30 glycerol-treated spectra (Fig 4). Additionally, the asymmetry of the illite 001 peak (Fig. 4) suggests that this expandable layer clay mineral is an illite-smectite mixed-layer phase (Moore and Reynolds Jr., 1997) and this is in agreement with the  Dypvik et al. (2011) of XRD spectra from core samples from the same formations exposed in the studied catchments. This mixed layer expandable phase will be referred to as 'I/S' in the following discussion. The relative proportions of illite, chlorite, kaolinite and I/S are given in Table 2.
The solid samples collected from Dryadbreen tend to have higher illite abundances and lower I/S abundances than those samples collected from Fardalen (Table 2). For all samples, there is an inverse relationship between the relative abundances of 5 I/S and illite (Fig. 4b). The relative abundances of kaolinite and chlorite were similar in both catchments (Table 2). We are not able to distinguish between authigenic and detrital clay minerals.

Sediment sources
The variation in clay mineralogy, 87 Sr/ 86 Sr and εNd 0 could be caused by two scenarios: weathering (either modern or in the 10 Eocene) or mixing between two or more sediment sources.
There are examples from previous studies where chemical weathering has been identified as the cause of an inverse correlation between the relative proportions of illite and smectite (Fig. 5b, e.g. Setti et al., 2004). However, whilst modern day weathering processes can induce large variations in 87 Sr/ 86 Sr primarily as a result of large inter-mineral variations in the Rb/Sr ratio (e.g. Bullen et al., 1997), it is much harder to induce large variations in the Sm/Nd ratio of minerals and this ratio is 15 often assumed to remain constant once a rock has been formed (e.g. McCulloch and Wasserburg, 1978). Fractionation of the Sm/Nd ratio during chemical weathering has been implicated in the generation of small εNd 0 offsets of around 2 epsilon units (Rickli et al., 2013) and larger variations in εNd 0 were observed in a soil profile developed on granitic till in northern Sweden (Öhlander et al., 2000). In that study a 7.7 epsilon unit variation was observed between the E horizon and the humic horizon, which was attributed to the preferential weathering of minerals enriched in Nd over Sm e.g. allanite. Nevertheless, in the given 20 time period of this study (50 Ma) it is not possible to generate a 14 epsilon unit variation without significant differences in mineralogy where individual minerals have different εNd 0 values. As there is no soil development in the studied catchments, the bulk mineralogy of the samples is broadly similar, as are their Sm/Nd ratios, we rule out chemical weathering as the primary cause of variation. Additionally, the major element chemistry is very similar to that observed in core samples drilled through the same formations (Fig. 2, Schlegel et al., 2013), confirming that weathering processes since the Paleocene have had 25 minor impact on bulk element and, by inference, 87 Sr/ 86 Sr and εNd 0 values. Therefore, the only way to generate the range of observed range in εNd 0 values is by the mixing of sources with distinct isotope ratios.
The linear trend between 87 Sr/ 86 Sr and εNd 0 is suggestive both of a common regional process affecting both isotope systems (Goldstein and Jacobsen, 1988) and of mixing between two end-members. The sediments deposited in the Central Basin during the Eocene were themselves derived from Mesozoic sediments. Based on zircon dating, it is thought that during the 30 Mesozoic, the sediment source to Svalbard alternated between an older (Proterozoic) western component comprised of reworked sediments from Greenland and Canada and a younger (Carboniferous-Jurassic) eastern component from Siberian foldand-thrust belts (Bue and Andresen, 2014;Elling et al., 2016). The erosion of the Siberian Traps which formed within the same time period (Permian to Triassic) would also have contributed sediment to the ocean. We will refer to these two sources as 'East' and 'West'.
The Eastern source (Lightfoot et al., 1993;Wooden et al., 1993;Spadea and D'Antonio, 2006) is relatively well-defined since the samples are essentially mono-lithologic (basaltic) and were deposited over a relatively short time-period. Based on zircon dating, both the Uralides and the Verkhoyansk Fold-and-Thrust Belt have been identified as potential sources to Svalbard 5 Mesozoic sediments and the Paleocene Basilika Formation (underlies the Grumantbyen Formation), respectively, (Bue and Andresen, 2014;Elling et al., 2016). The western end-member is much harder to characterise as it consists of Archaean rocks which have undergone extensive metamorphism. We require an end-member with εNd 0 values lower than -24.2 (R03, Table   1) and we therefore only consider data from western Greenland as the East coast was affected by the Caledonian orogeny (Henriksen, 1999) and later by the rifting of the North Atlantic (Bernstein et al., 1998) and therefore has higher (younger) 10 εNd 0 values (Jeandel et al., 2007). The range in 87 Sr/ 86 Sr and εNd 0 from literature data of Archaean rocks from western and northern Greenland (Jacobsen, 1988;Collerson et al., 1989;Weis et al., 1997;Kalsbeek and Frei, 2006;Friend et al., 2009) is 0.70153 to 2.33356 and -56 to -2.75. By changing the Sr/Nd ratio of these two end-members, mixing lines can be drawn which encompass all the data, with the majority of points falling on a mixing lines with an r value of 1 (i.e. the Sr/Nd ratio of both end-members is the same, Fig. 6).

15
The variation in clay mineralogy (Fig. 5) can be explained by the different lithological sources of the two end-members ( Fig.   6). Basalt typically weathers to smectite group minerals (e.g. Curtin and Smillie, 1981;Parra et al., 1985) and modern sediments originating from Siberia (basaltic) are enriched in smectite (Nürnberg et al., 1994;Wahsner et al., 1999). Any volcanic particles present will also tend to weather to smectite (Bayon et al., 2015). The western source is dominated by granitic rocks where the mica and K-feldspar typically weather to illite and kaolinite, respectively (Essington, 2004). As the western source is older, 20 illite has high Rb/Sr ratios and detrital illite is resistant to weathering, this results in high 87 Sr/ 86 Sr and low εNd 0 . By contrast the younger eastern source will have lower 87 Sr/ 86 Sr and higher εNd 0 values. These distinct differences between the two sources leads to the observed correlations between clay mineralogy, 87 Sr/ 86 Sr and εNd 0 values (Fig. 5). Schlegel et al. (2013) concluded that, on the basis of microscopic observations, the geochemical changes observed between the different formations arose as a result of increased chemical weathering during the late Paleocene and not as a change in 25 source rock provenance, which remained from the west. However, that western source in the Paleocene-Eocene was itself comprised of two sources deposited in the Mesozoic. Chemical weathering during the Paleocene cannot be reconciled with the wide variation in 87 Sr/ 86 Sr and εNd 0 values. Therefore, the overall trend observed between 87 Sr/ 86 Sr and εNd 0 is more likely caused by two 'proto-sources' of different ages mixing: an illite-rich end-member with high 87 Sr/ 86 Sr and low εNd 0 and an illite-poor end-member with low 87 Sr/ 86 Sr and high εNd 0 .

Difference between catchments
Suspended sediments from the glaciated catchment (Dryadbreen) are distinct from the suspended sediments from the unglaciated catchment (Fardalen) with lower εNd 0 , higher 87 Sr/ 86 Sr values and a greater relative proportion of illite. This is consistent with the rock and sediment samples collected from the two catchments (Tables 1 and 2). Previously published clay mineralogy and major element data was used to determine which formations the collected rock and sediment samples likely originated from. In a study on the clay-sized fraction of core samples from the Gilsonryggen Member (Riber, 2009;Dypvik et al., 2011), which is a member of the Frysjaodden Formation observed to the south-west of the Central Basin overlying a sandstone wedge (the Hollenderdalen Formation), the relative proportion of illite varied from 21-36% and mixed-layer phases were detected, accounting for 17-35% of the clay minerals, in agreement with data from R01 5 and G (illite 36-46%, I/S 40-50%), suggesting that these samples are from the Frysajodden Formation. The major element chemistry of these two rock samples is also consistent with core data from the Frysjaodden Formation (Schlegel et al., 2013;Hindshaw et al., 2016). Additionally, Schlegel et al. (2013) report the relative proportions of illite, chlorite and smectite in core samples from Aspelintoppen, Battfjellet and Frysjaodden formations. There is a decrease in illite content from 64% in Aspelintoppen to 51% in Frysajodden, suggesting that the samples with a high relative proportion of illite (e.g. D, Table 2) are 10 derived from the Aspelintoppen Formation.
The changes in clay mineralogy between formations are caused by changes in the deposition environment of the Central Basin: from a pro-delta environment (Frysjaodden Formation) to delta progradation (Battfjellet Formation) to finally being filled (Aspelintoppen Formation) (Müller and Spielhagen, 1990;Helland-Hansen, 2010). Sediments are sorted as a function of particle size as they travel through the water, such that coarser particles (typically primary minerals such as feldspar and quartz) 15 will settle faster than finer particles (clay minerals) and within the clay minerals, illite will settle faster than smectite (Sionneau et al., 2008). A size-sorting effect is observed in the difference between the 87 Sr/ 86 Sr and εNd 0 values of the bulk and the clay-sized fraction. This effect is observed at a global scale and is interpreted to reflect the preferential transport of volcanics and basalt in the fine fraction (Bayon et al., 2015). A similar principle could contribute to the observed array in 87 Sr/ 86 Sr-εNd 0 space (Fig. 3). The Frysjaodden Formation, being furthest away from shore, became enriched in smectite-enriched particles 20 derived from the basaltic eastern end-member whereas the Aspelintoppen Formation, deposited in a near-shore environment, became enriched in coarser illite-enriched particles derived from the granitic western end-member. Furthermore, smectite is the only clay mineral which forms in significant amounts in seawater (Griffin et al., 1968) and therefore it is very likely that the deep-sea Frysjaodden Formation contains authigenic smectite in addition to smectite derived from continental weathering, increasing the relative proportion of smectite in this Formation. Taken together, this would suggest that the variation in clay 25 abundances and isotope ratios we observe in the solid samples could reflect the greater relative proportion of the Aspelintoppen Formation relative to the Frysjaodden Formation in rocks and sediments collected in Dyradbreen compared to Fardalen. This would be consistent with the moraine material being predominantly derived from rocks once located in the upper reaches of the catchment (Aspelintoppen Formation) and the modern day sandur plain, containing the products of this physical erosion, is essentially burying the lower down Frysajodden Formation. The difference in the suspended sediments from the two catchments

Effect of leaching
The chemical and isotopic composition of leached suspended sediment is distinct from bulk suspended sediment (Fig. 7a). The residual phase is depleted in MREE (middle rare earth element) and the greater the MREE depletion, the greater the difference in εNd 0 between residue and bulk (Fig. 7b). From mass balance constraints, this points to the existence of a labile pool with low 87 Sr/ 86 Sr and high εNd 0 , which is MREE enriched. There are several possibilities for what the leached phase could be 5 (volcanics, carbonate, apatite) and we will consider each below.
The labile phase could be volcanics. Volcanic ash has a MREE enriched REE pattern (Tepe and Bau, 2014) and would have high εNd 0 and low 87 Sr/ 86 Sr. However, the amount of a volcanic component is expected to be minor in the studied sediments as the volcanic ash component of particulates readily leaches upon contact with seawater (Pearce et al., 2013;Wilson et al., 2013) and therefore may already have been leached during deposition in the Paleocene-Eocene. Additionally, volcanic ash in 10 these layers has been diagenetically altered to bentonites (Cui et al., 2011;Elling et al., 2016;Jones et al., 2016) which are unlikely to be readily leached.
Carbonate minerals would be expected to readily leach in 5% HCl. However, carbonates are unlikely to be the dominant phase leaching because carbonate minerals were only detected in sediments from Dryadbreen (Fig. 4, Hindshaw et al., 2016).
The sediments from Dryadbreen would therefore be expected to show the greatest offset between bulk and residue, but in fact 15 it is the sediments from Fardalen, which contain no carbonate, which exhibit the greatest offset between bulk and residue (Fig.   7b). Additionally REE in carbonates are typically LREE enriched relative to chondrite (as are shales) and often have Ce and Eu anomalies (Shaw and Wasserburg, 1985;Schieber, 1988;German and Elderfield, 1990;Komiya et al., 2008;Hua et al., 2013).
The residue in Fig. 7b would therefore not be expected to have such a pronounced MREE depletion relative to the bulk, and no obvious Ce or Eu anomalies are observed (Fig. 7b). Finally, carbonate minerals do not typically contain high concentrations of 20 REE (Shaw and Wasserburg, 1985) and it is more likely that other accessory phases, such as apatite, are the main REE carriers.
The REE composition of apatite depends on its origin. Leaching of crystalline apatite leads to a LREE depletion of the residue compared to the bulk in granitic/gneissic catchments (Aubert et al., 2001;Négrel, 2006) and is unlikely to contribute to the leached phase in this study. In contrast, a 'humped' REE pattern in a leachate (i.e. the reverse of Fig. 7b) is characteristic of diagenetically altered apatite (Tricca et al., 1999). Further studies have also linked MREE enrichment in the leached fraction 25 to diagenetic changes associated, not just with phosphates, but also with Fe/Mn oxyhydroxides and clays (Ohr et al., 1994;Johannesson and Zhou, 1999;Su et al., 2017). A diagenetic phase is also consistent with the isotope data where the leached phase is 'younger' (high εNd 0 and low 87 Sr/ 86 Sr) compared to the residue.
The shales in the Frysjaodden Formation were deposited in the marine environment and any authigenic minerals which were formed at that time are likely to have incorporated fluids with the Eocene seawater composition and deep-sea clays are most 30 susceptible to incorporating seawater (Dasch, 1969). This is consistent with the Fardalen sediments, which contain more of the deep-sea Frysjaodden Formation (see previous section), being more affected by leaching than the Dryadbreen sediments ( Fig. 7b). In addition to diagenetic changes, adsorption may also occur. Samples containing a greater relative proportion of I/S, have a greater cation exchange capacity and are therefore more likely to contain a greater proportion of ions from seawater, increasing the difference between 87 Sr/ 86 Sr and εNd 0 in the residue and bulk (Dasch, 1969;Ohr et al., 1991). The greater difference between εNd 0 in bulk and residue in the Fardalen sediments compared to Dryadbreen is therefore consistent with the greater proportion of I/S in the Fardalen sediments. Adsorption of Nd from seawater was also implicated in a study by von Blanckenburg and Nägler (2001) where leachates of marine sediments had higher εNd 0 than bulk and terrestrial sediments which had had no contact with seawater showed the reverse pattern (lower εNd 0 in leachate compared to bulk).

5
Assuming the leached phase is comprised of a mixture of authigenic minerals such as apatite and cations readily leached from clay minerals, then the leachate should have a seawater isotopic composition. Radiogenic Sr in seawater in the past is relatively well constrained given that it has a uniform value across the worlds oceans. Radiogenic neodymium, on the other hand, varies between ocean basins. A study based on fish, provides some constraints on the εNd 0 and 87 Sr/ 86 Sr isotopic composition of the Arctic Ocean during the Eocene (Gleason et al., 2009), with 87 Sr/ 86 Sr varying from 0.7078 to 0.7088 and εNd 0 varying 10 from -7.5 to -5.5. If we assume that the location of this study and the area of the future Central Basin were connected, then this end-member would be within error of the eastern end-member and therefore could not be distinguished (Fig. 6). This is the most likely reason why the two trends (residue-bulk-leachate and east-west bulk) appear to fall on a common mixing line (Fig.   7a). However, the potential for modification of the main mixing trend set by rock type (Fig. 6) by later diagenetic processes cannot be ruled out and would support the conclusions of Awwiller (1994) who concluded that provenance information based 15 on Nd-Sr isotopes could be obscured by the partial incorporation of Sr and Nd from seawater. Diagenetic alteration has been implicated in shales which give unrealistically old Nd model ages (Arndt and Goldstein, 1987;Awwiller and Mack, 1991;Bock et al., 1994;Cullers et al., 1997;Krogstad et al., 2004).

Implications for Nd as a sediment source tracer
The sediments observed in this study have highly heterogeneous Nd isotopic compositions and the difference in suspended 20 sediment load between the two catchments is up to 6.8 epsilon units. Additionally, seasonal variation is observed: 0.6 epsilon units in Dryabreen and 1.6 epsilon units in Fardalen. A similar magnitude of seasonal variation in εNd 0 has previously been reported in the sediments of much larger rivers. A 1.3 seasonal variation has been reported in suspended sediments from the Madeira River (Amazon, Viers et al., 2008) and a 2 epsilon unit range was observed sediments from two tributaries of the Ganges (Kosi and Narayani, Garçon et al., 2013). The seasonal variation in both of these studies was attributed to the seasonal For the purposes of using Nd in ocean sediment cores as a tracer for past sediment sources it is assumed that the Nd isotopic composition of sediments is constant for broad source regions (e.g. Jeandel et al., 2007), and this will not be affected by sea- sonal variations. However, seasonal cycles give an insight into weathering and erosion conditions under different hydrological regimes that are an analogue for longer term trends. It is entirely plausible that an intensified or weakened hydrological cycle could change the Nd isotopic composition of sediment export for a given region. Of particular relevance to the Arctic region is the re-organisation of drainage basins as the ice sheets waxed and waned and the attendant changes in magnitude and location of discharge to the ocean (e.g. Teller, 1990;Wickert, 2016). Therefore, it should not necessarily be assumed that the continental 5 regions have had a constant εNd 0 export to the oceans. For example, the 5.7 epsilon unit range in εNd 0 (which is of similar magnitude to difference between catchments observed in this study) in an Arctic sediment core (PS1533, Tütken et al., 2002) over the last 140 ka was attributed to changes in the relative proportion of sediment derived from two isotopically distinct sources (Svalbard and Siberia) over glacial-interglacial cycles. Although the broad end-member identification will unlikely be affected, the calculated proportions of each end-member at different points in time would change if the those end-members are 10 not constant over glacial-interglacial cycles.

Conclusions
The large variations in 87 Sr/ 86 Sr and εNd 0 observed in two small catchments in Svalbard can be explained as a result of two isotopically and geochemically distinct 'proto-sources' mixing during the Meoszoic and subsequently forming the Paleocene-Eocene sedimentary formations which are eroding today (Fig. 8). The two original sources are an eastern source derived from 15 basaltic rocks from Siberia and a western source derived from granitic rocks from Greenland. The original geology of the sources controls the initial geochemistry, Sr and Nd isotope values and subsequently determines the type of clay minerals formed during weathering, susceptibility to later diagenesis and particle-size transport effects. Changes in erosion caused by the glaciation of Dryadbreen has led to material from the upper (younger) formation, which contains a higher proportion of material derived from the western source, being moved lower down in the catchment where 20 it is present in the moraines. In contrast, the lower (older) formation, which contains a higher proportion of material derived from the eastern source, is fully exposed in the unglaciated catchment, having not been covered by sediment from the upper reaches of the catchment. This leads to a marked difference in the suspended sediment export from the two catchments and suggests that changes in continental erosion during glacial-interglacial cycles could have a pronounced effect on the Sr and  supraglacial sediment (yellow circle), Fardalen (blue circle) and Dryadbreen (red circle). Other rock and sediment samples were collected at various locations within the two catchments ( Table 2). The red dot in the inset shows the location of the study area (Latitude, 78 • 08'N; Longitude, 15 • 30'E) in relation to the rest of Svalbard. Geological information is taken from (Major et al., 2000) (001) illite (002) illite (003) quartz kaolinite (002) chlorite (004) kaolinite (001) chlorite (002) chlorite (003)  illite (001) illite (002) illite (003) kaolinite (002) chlorite (004) kaolinite (001) chlorite (002) chlorite (003) chlorite ( (Lightfoot et al., 1993;Wooden et al., 1993) and Uralides (Spadea and D'Antonio, 2006). 'West' data (n=65): Archaean rocks (predominantly gneisses, Jacobsen, 1988;Collerson et al., 1989;Kalsbeek and Frei, 2006;Friend et al., 2009), basal ice particles from GISP 2 and GRIP and granite bedrock samples from GISP 2 (Weis et al., 1997). The 'r' values are the Sr/Nd ratio of the 'East' source divided by the Sr/Nd ratio of the 'West' source. The star indicates the isotopic composition of Eocene seawater (see text for details, Gleason et al., 2009 Table A1 2 Samples names are YYYYMMDD and the subsequent letters are D=Dryadbreen (glaciated), F= Fardalen (unglaciated) and SG = supraglacial  The hotplate digestion method used does not digest zircons and therefore the high rare earth element (HREE) concentrations may be lower than the true total. Nd and Sm are fully digested by the hotplate method (Rickli et al., 2013).