Introduction

Changes in Atlantic Meridional Overturning Circulation (AMOC) modulate orbitally-forced glacial to interglacial climate change by altering the distribution of heat and carbon storage in the deep sea1. Reconstructions of AMOC configuration indicate more sluggish circulation during glacial periods when North Atlantic Deep Water (NADW), the northern-sourced end member formed by convection in the Nordic and Labrador seas, shoals to intermediate depths and is replaced by southern-sourced waters extending into the deep North Atlantic2,3,4,5,6,7,8. Reconstructions of past circulation are often based on records of Nd isotopes preserved in authigenic phases, such as Fe-Mn oxide coatings on sediments or foraminifera, that are considered to be seawater archives. Nd isotopes are assumed to be conservative tracers of water masses because Nd has a short residence time relative to the global ocean mixing time. Accordingly, water masses carry distinct Nd isotopic signatures (reported as ɛNd, the 143Nd/144Nd ratio normalized to chondrite and multiplied by 104) that originate from differences in values of proximal continental sources controlled by bedrock age and lithology. Due to this conservative behavior, seawater ɛNd variations have been attributed to changes in mixing proportions of end member water masses9,10,11,12. However, there is increasing recognition that ɛNd of intermediate and deep waters may be altered by non-conservative seawater-particle interactions within the water column13,14 and/or at the sediment-water interface15,16,17,18,19. Thus, authigenic ɛNd lower than any major modern water mass recorded in the North Atlantic during glacial terminations (Fig. 1) has been attributed to increased production of Baffin Bay Water (BBW, ɛNd ~−2020) and Labrador seawater (LSW, ɛNd ~−13.915)5,6,7, or to nonconservative release of unradiogenic Nd from fresh shield sediments delivered to the North Atlantic during retreat of Northern Hemisphere ice sheets21,22,23,24. In this paper, we test the hypothesis that weathering extent of shield-derived detrital inputs controls the long-term magnitude and direction of ɛNd variations in North Atlantic deep water.

Fig. 1: Nd isotope records from the western North Atlantic for the last glacial cycle.
Fig. 1: Nd isotope records from the western North Atlantic for the last glacial cycle.
Full size image

a Authigenic (circle = sediment leachate; square = foraminifera) and detrital (triangle) ɛNd records from Baffin Island Shelf (BIS, 2674 m depth, green), Orphan Knoll (OK/Site U1302, ~3500 m depth, purple), Bermuda Rise (BR/Site 1063, ~4550 m depth, pink), and Blake Bahama Outer Ridge (BBOR/Site JPC12, ~4250 m depth, red). BIS leachates, and BIS and OK foraminifera and detrital data from69; OK leachates from70; BR foraminifera data from5; BR detrital data from48; BBOR leachates from23; and BBOR detrital data from71. Error bars represent two standard deviations (2 SD). The gray bars correspond to cold periods of Younger Dryas (YD) and Heinrich Stadials 1 (H1) and 2 (H2). The blue bar on the y-axis shows values of modern lower NADW17. b Location map of core sites, the Labrador Sea (LS), the Deep Western Boundary Current (DWBC), the sample transect in SW Greenland (fuchsia rectangle), and the cross section for Fig. 6 (black line). Map data ©2015 Google.

Leaching experiments and field observations suggest a mechanism by which chemical weathering of comminuted Precambrian material contributes Nd with low ɛNd to North Atlantic seawater during deglaciations. Nd leached from stream sediment derived from Archean rocks is up to 17 ɛNd units lower than bulk material25, and Greenlandic stream water has been observed to be 4 ɛNd units lower than co-located suspended sediment26. However, little work has focused on the weathering kinetics of these glacial sediments derived from Precambrian shield terrains to understand the processes allowing the transfer of unradiogenic Nd from the land to the sea. To address this question, we assess changes in mineral weathering and ɛNd of paired sediments and stream waters across a 170 km transect extending from the Greenland Ice Sheet (GrIS) to the coast in SW Greenland. In response to deglaciation following the Last Glacial Maximum (LGM), exposure ages in this region range from 0 at the ice edge to 12 ky at the coast27 (Fig. 2a). Systematic shifts in co-located dissolved and particulate ɛNd across this Precambrian bedrock transect indicate that Nd isotope exports change as landscapes evolve in response to winnowing and differential mineral weathering, and illustrate how these processes could contribute to variations in North Atlantic seawater and detrital ɛNd on glacial/interglacial timescales.

Fig. 2: Nd isotopes of stream water and sediment across the coast to ice sheet transect in SW Greenland.
Fig. 2: Nd isotopes of stream water and sediment across the coast to ice sheet transect in SW Greenland.
Full size image

a Location of sampled watersheds. Akuliarusiarsuup Kuua/Watson River complex (AKR/WR, yellow) is a proglacial watershed. Isunngua (purple), Lake Helen (red), and Qorlortoq (orange) near Kangerlussuaq, and Nerumaq (blue) and Sisimiut (cyan) near Sisimiut are deglaciated watersheds. All the sites are in the Nagssugtoqidian Mobile Belt (NMB). Location of the Archean Block (AB) is highlighted in light brown. Geological provinces from the Greenland Mineral Resources Portal (greenmin.gl) of the Geological Survey of Denmark and Greenland and the Ministry of Mineral Resources under the Government of Greenland. Moraine exposure ages (white dashed lines) vary from ∼12 ka at the coast to 0 ka at the Greenland Ice Sheet (GrIS)27. The water balance (orange dashed line) shifts from positive closer to the coast to negative closer to the ice72. Brackets at the top of panel a indicate the spatial extent of watersheds. Map data ©2015 Google. b. Boxplots of dissolved load ɛNd (brown-outline), bedload ɛNd (black-outline), suspended sediment ɛNd (gray-filled), and respective average ɛNd (colored symbols) for each watershed. Numbers at whisker ends indicate the number of samples.

Results and discussion

ɛNd variations across the transect

Watersheds in the coast-to-ice transect in SW Greenland are underlain by Archean orthogneisses and Paleoproterozoic granitoids metamorphosed to amphibolite to granulite facies during the Proterozoic. These rocks are tonalitic to quartz dioritic in composition28. A systematic decrease in the offset between stream water and bedload Sr isotope ratios toward the coast indicates a progressive increase in sediment weathering extent with exposure age29,30, which drives changes in solute composition, nutrient export, and reactivity of dissolved organic carbon29,30,31,32. We analyzed Nd isotopes in co-located bedload sediment and stream water from multiple sites within a proglacial watershed (Akuliarusiarsuup Kuua/Watson River; AKR/WR), which is primarily sourced by glacial meltwater, and five deglaciated watersheds that lack glacial meltwater and have variable exposure ages (Fig. 2b). Bedload ɛNd values across the transect reflect varying contributions of Archean and Proterozoic bedrock sources, with average values increasing from ~−31 near the ice to ~−28 at the coast. Average dissolved ɛNd shows a larger shift, increasing from ~−39 in ice-proximal watersheds to ~−29 in coastal watersheds. As a result, the ɛNd offset between bedload and dissolved phases (ɛNdsediment - ɛNddissolved) decreases from ~8 ɛNd units near the ice to ~1 unit near the coast. These trends indicate preferential dissolution of less radiogenic Nd in the most recently exposed watersheds and more congruent Nd isotopic release in watersheds with older exposure ages.

Analyses of grain size separates from one representative bedload sample per watershed also show systematic shifts in grain size distributions, Nd concentrations, and ɛNd across the transect. The lowest ɛNd occurs in clay size fractions, with higher values in silt and then sand size fractions (Fig. 3). Silt and clay ɛNd values increase toward the coast, where they converge to align with sand and bulk bedload ɛNd. Clay and silt abundances decrease from the proglacial watershed, where they represent 1.2 and 48.6 wt%, respectively, to the coast, where combined they constitute less than 1.5 wt% of the bedload (Fig. 4a), reflecting a system in which fluvial processes have removed most of the glacial flour, mainly composed of silt-sized particles, over time. In the proglacial AKR/WR and the two ice-proximal deglaciated watersheds, Isunngua and Lake Helen, silt represents ~30–50 wt% of the bedload but carries ~55–70% of the Nd mass budget based on ɛNd and Nd concentrations of the grain size fractions (Fig. 4b). In the coastal watersheds, silt is ~1 wt% of bedload sediment and constitutes ~3% of the Nd budget, which is dominated by the sand fraction. These findings suggest that minerals enriched in unradiogenic Nd tend to be concentrated in fine sediments and that the decreasing proportion of silt toward the coast contributes to the increase in bulk bedload ɛNd. These transitions in particulate and dissolved Nd isotopes from near-ice to coastal watersheds reveal changes in sediment transport and weathering reactions with exposure age. Specifically, both processes are approaching equilibrium with local environmental conditions33, similar to geomorphic landforms34,35, Sr isotopes36, and greenhouse gases37.

Fig. 3: ɛNd and Nd concentrations of grain size separates across the SW Greenland transect.
Fig. 3: ɛNd and Nd concentrations of grain size separates across the SW Greenland transect.
Full size image

ɛNd and Nd concentrations (indicated by symbol size) of clay (<2 µm; pink hexagons), silt (2–63 µm; green squares), and sand (63 µm–2 mm; blue diamonds) from one sample for each watershed against distance from the ice. Also included are bulk bedload (black triangles) and average suspended sediment (downward-pointing gray triangles) from AKR/WR (n = 73) and Sisimiut (n = 21). Brackets at the top of the figure indicate the spatial extent of watersheds. Grain size and bulk bedload symbols are centered on the sample location within the watershed.

Fig. 4: Grain size distributions and contributions to bedload Nd budget across the SW Greenland transect.
Fig. 4: Grain size distributions and contributions to bedload Nd budget across the SW Greenland transect.
Full size image

a Weight proportions of clay (solid pink), silt (striped green), and sand (dotted blue) size separates for one sample from each watershed against distance from the ice. b Nd contributions from each grain size to the bulk bedload Nd budget against distance from the ice. Brackets at the top of the figure indicate the location of watersheds. Bars are centered on the sample location within the watershed.

Effect of weathering kinetics on Nd isotopes

Changes in mineral weathering across the SW Greenland transect have been interpreted from Sr isotopes29,30. Higher Sr isotopic ratios in stream water compared to bedload in newly exposed watersheds were attributed to the preferential release of radiogenic Sr from interlayer sites in biotite, while similar ratios in watersheds with longer exposure ages were associated with more congruent dissolution of Sr isotopes during chemical weathering of plagioclase38. In contrast to the Rb-Sr system, the parent and daughter isotopes in the Sm-Nd system have similar geochemical characteristics, so fractionation of Nd isotopes during weathering is expected to be negligible. However, minerals crystallize with distinct Sm/Nd ratios, which produces mineral-specific ɛNd as Sm decays (Fig. 5). As a result, minerals with a lower resistance to weathering may release Nd with ɛNd that differs from that of the bulk mineral assemblage. For example, preferential weathering of reactive mineral phases with low ɛNd, such as allanite, would produce large offsets between bedload sediment and dissolved ɛNd during early stages of weathering, as observed in the proglacial and near ice sheet deglaciated watersheds (Fig. 2b). Preferential dissolution of allanite as a primary source of unradiogenic dissolved ɛNd in ice-proximal watersheds is supported by allanite ɛNd as low as −46 in Qorlortoq39. Allanite was also identified as the phase dominating the Pb isotope and REE compositions of initial experimental leachates from the youngest soils (~12 ka) in a granitoid chronosequence in Wyoming40.

Fig. 5: Relative ɛNd and Nd concentrations of minerals present in SW Greenland.
Fig. 5: Relative ɛNd and Nd concentrations of minerals present in SW Greenland.
Full size image

Relative ɛNd (upper x-axis) based on typical ratios of partition coefficients for Sm and Nd (DSm/DNd - lower x-axis). Nd concentrations (grayscale) range from a few ppm up to a few thousand ppm, with darker tones indicating higher concentrations. Minerals are listed by relative weathering resistance (y-axis). Based on Bayon et al.73 with information from references in the Supplementary Note 1.

Preferential chemical weathering of REE-bearing minerals with low ɛNd may be enhanced by their high abundance in fine sediment fractions39 that have elevated surface area to volume ratios. Furthermore, similar ɛNd (Fig. 3) and REE patterns (Supplementary Fig. 1) of the silt fraction of bedload sediment and suspended sediment in AKR/WR suggest that export of silt as suspended load also contributes to the selective removal of these REE-enriched minerals (Fig. 4). A decrease in the amount of silt-sized minerals with low ɛNd and an increase in relative contributions from more radiogenic and weathering-resistant major rock-forming minerals such as amphiboles and pyroxenes (Fig. 5) would result in a shift from incongruent to congruent release of Nd isotopes, decreasing the offset between bedload and dissolved ɛNd with increased exposure age (Fig. 2). This shift from weathering of REE-enriched trace minerals to major rock-forming minerals would also reduce the overall release of Nd and other REEs. These results highlight how weathering of shield-derived sediments could impact Nd export and provide insights for understanding marine records of Nd isotopes.

Weathering and North Atlantic ɛNd record

Glacial terminations expose large volumes of comminuted sediments from equivalent Precambrian terranes in Eastern Canada and SW Greenland. Enhanced chemical weathering of these reactive sediments has been interpreted to cause unradiogenic Nd21,22,23,24 and radiogenic Pb41,42,43,44 excursions in North Atlantic authigenic phases during glacial-interglacial transitions, but processes and chemical reactions in both terrestrial and marine settings that control isotopic variations have not been investigated. Our results confirm that seawater-sediment interactions, as influenced by the reactivity and supply of comminuted shield-derived sediments, contribute to North Atlantic deep water ɛNd15,17,21,22,23,24. In addition, deep water ɛNd will depend on AMOC strength, which determines the seawater-sediment interaction time and the transport of solutes and sediments from upstream deep water sources19,45,46. Documented changes in Nd isotopes across the exposure age and weathering gradient in SW Greenland watersheds highlight the critical role of weathering extent in sediment reactivity and can improve interpretations of seawater ɛNd variations through time. In general, shield-derived unweathered and fine-grained glacial sediment exported to the ocean during deglaciation should weather rapidly and preferentially release less radiogenic Nd, similar to our observations in ice-proximal watersheds (Fig. 2). In contrast, prolonged weathering in the absence of fresh, reactive sediment inputs should lead to congruent release of more radiogenic Nd, as observed in our coastal watersheds. A similar weathering extent-related mechanism could also produce the radiogenic Pb excursions recorded in authigenic records during periods of elevated sediment inputs as shown by decoupling of dissolved and bedload Pb isotope ratios in the ice-proximal streams (Supplementary Fig. 2). However, potential contamination of stream water by anthropogenic Pb, particularly in deglaciated watersheds with more extensive weathering, limits the ability to evaluate the effect of progressive weathering on Pb isotopes across our transect.

Multiple processes may shape and modify the record of Nd isotopes in authigenic phases47,48,49. First, ɛNd of authigenic phases can reflect varying contributions of pore water and seawater. Furthermore, potential detrital contamination during laboratory extraction of Nd by leaching may bias the measured authigenic ɛNd. Given the ubiquity of Fe-Mn oxyhydroxide coatings and their high Nd concentrations compared to other phases such as foraminifera, multiple leaching protocols using various reagents and reagent strengths have been developed to target these coatings while minimizing detrital contamination49,50,51,52. Notably, in some sites, different leaching techniques produce inconsistent results, likely reflecting varying degrees of dissolution of different reactive detrital phases48,52. However, shifts in the Nd isotopic composition of leachates and their offset from detrital εNd provide information about changes in sediment sources and sediment reactivity, offering insights into the amount and isotopic signature of Nd potentially dissolved at a given time. Below, we leverage the spatial and temporal coverage of sediment leachate ɛNd analyses and the more limited detrital ɛNd data in the western North Atlantic (Fig. 1) to determine how the export and weathering extent of shield-derived sediment during the Laurentide Ice Sheet retreat across the Canadian Shield may have influenced the εNd of the North Atlantic deep water (Fig. 6). Only leachates extracted using the protocol described by Blaser et al50. are considered to minimize variations introduced by differential extraction from distinct detrital phases and to allow direct comparison across sites and through time.

Fig. 6: Conceptual model illustrating the effects of shield-derived sediment inputs and weathering extent and AMOC strength on Nd isotopes in the Labrador Sea throughout the last glacial cycle.
Fig. 6: Conceptual model illustrating the effects of shield-derived sediment inputs and weathering extent and AMOC strength on Nd isotopes in the Labrador Sea throughout the last glacial cycle.
Full size image

Colored arrows indicate the strength (size) and ɛNd (color) of the DWBC entering from the Nordic Seas (right) and exiting the Labrador Sea (left). ɛNd of Nordic Seas input from70. The size of white arrows indicates the flux of Nd from weathering reactions at the sediment-water interface. The colored portion of the ocean represents the ɛNd of lower NADW. Figure not to scale. a LGM: Limited sediment influx due to ice on proximal continents and sea ice on the LS, and a sluggish NADW deep circulation, would lead to reduced weathering reactions. b H1: Ice rafting and gravity flows of unweathered shield material would enhance weathering reactions. A weakening of AMOC would limit additional imports and exports of Nd. c Early deglaciation: Delivery of chemically mature shelf sediment as the LIS retreated across the continental shelf would decrease weathering fluxes. A strengthening DWBC would import radiogenic Nd and export less radiogenic Nd. d Early Holocene: Input of freshly comminuted shield-derived sediment as the LIS retreated across continental bedrock would enhance weathering fluxes. A strong DWBC would continue importing radiogenic Nd and exporting considerably less radiogenic Nd. e Present: Limited fresh sediment inputs due to the loss of the LIS decreased weathering fluxes. A strong DWBC imports radiogenic Nd and exports relatively less radiogenic Nd.

During the LGM, limited influx of glacial sediment to the North Atlantic53 combined with weak NADW deep circulation2,3,4,5,6 would have allowed bottom waters to approach chemical equilibrium with seafloor sediments resulting in minimal dissolution of relatively radiogenic Nd at Baffin Island Shelf (BIS) and Orphan Knoll (OK) (Fig. 6a). The leachate and detrital ɛNd of ~−18 (Fig. 1) suggest deep water interactions with weathered sediments derived from Precambrian and Paleozoic rocks, likely deposited during Heinrich Stadial 2 (H2), with little input from younger Iceland and northwest European sources given the reduced AMOC. During Heinrich Stadials 2 (H2) and 1 (H1), abundant unweathered, glacially comminuted shield-derived sediment was delivered to the subpolar North Atlantic via gravity flows and ice-rafting through Hudson Strait53. Reactions between seawater and this rapidly deposited unweathered sediment with low ɛNd (~−27 to −23) during an interval of sluggish AMOC4,6,54 would have led to the release of Nd with ɛNd lower than detrital values (~−30 to −24) at BIS and OK (Fig. 6b), while seawater and detrital ɛNd would remain unchanged at Bermuda Rise (BR) and Blake Bahama Outer Ridge (BBOR) (Fig. 1). As the ice sheet retreated across the continental shelf during early deglaciation (~16–13 ka), partially weathered shelf sediment would have been transported to the deep sea, increasing seawater and detrital ɛNd in BIS and OK (Fig. 6c). In addition, a strengthening of the Deep Western Boundary Current (DWBC) at this time would have enhanced the transport of dissolved and particulate Nd. Relatively radiogenic Nd from Icelandic and Nordic sources would have been carried into the Labrador Sea (LS), with the strongest influence at the most proximal site (BIS), while less radiogenic Nd from the LS was exported southward, decreasing ɛNd at BR and BBOR. Authigenic ɛNd generally higher than detrital ɛNd at this time suggest that young reactive phases, likely sourced from Iceland55, dominated the dissolved Nd budget.

During the early Holocene (~12–8 ka), ice sheet margins shifted from the continental shelf onto land56, leading to increased abrasion of continental bedrock and transport of freshly comminuted shield-derived sediment to the North Atlantic via proglacial streams and gravity flows driven by meltwater pulses21,53 (Fig. 6d). Chemical weathering of these reactive sediments in the deep sea would release unradiogenic Nd that, along with the unradiogenic particulate Nd, would be transmitted across the western North Atlantic by the reinvigorated DWBC, as evidenced in the steep decrease in leachate and detrital εNd (Fig. 1). Although this shield-derived sediment likely had εNd similar to the material exported during H1 and H2, higher detrital and leachate εNd during the early Holocene reflect important contributions from more radiogenic upstream sources along the DWBC at this time. Subsequent ice sheet retreat across terrestrial landscapes over the next 8 ky would lead to the development of deglaciated streams yielding low amounts of relatively radiogenic weathered sediment while newly comminuted unradiogenic sediment exported in proglacial streams would be trapped in fjord systems. Limited shield-derived sediment input and continued introduction of dissolved radiogenic Nd from younger, Icelandic and Nordic sources, transported by an active AMOC, would increase and homogenize seawater ɛNd throughout the Holocene (Fig. 6e). This is consistent with a study of variations in leachate ɛNd with distance from the LS that suggests the ɛNd of bottom waters advected across the western North Atlantic increased by ~1.4 units since the early Holocene23.

This analysis illustrates that marine authigenic and detrital ɛNd records are reliable tracers of weathering processes linked to glacial retreat and provide insights into changes in ocean circulation and global climate.

Implications

Shifts in stream water and bedload ɛNd associated with chemical weathering extent across the SW Greenland glacial foreland (Fig. 2) support the hypothesis that the ɛNd of LSW, and therefore NADW, respond to variations in the extent of weathering of glacial shield-derived sediment exported during ice sheet retreat. The magnitude and direction of the changes in ɛNd are largely controlled by the reactivity, Nd concentration, and Nd isotopic composition of mineral suites of this sediment. Considering that the most weatherable phases (e.g., allanite, apatite, titanite) within Precambrian terranes are enriched in Nd (Fig. 5) and REEs in general, the amount of Nd released should be higher during initial chemical weathering and decrease with weathering extent. Thus, high chemical weathering rates of reactive minerals in the sediment produced and transported to the seafloor during early deglaciation of terrestrial landscapes would lead to increased release of Nd with low ɛNd and other REEs into North Atlantic deep water. Progressive weathering of terrigenous material in the deep ocean during full interglacial and glacial conditions would reduce mineral dissolution, lowering concentrations of Nd and increasing ɛNd in the seawater. Paleocirculation models based on the distribution of authigenic Nd isotopes as a proxy for seawater should incorporate changes in Nd isotope loading to North Atlantic deep water to prevent overestimating LSW production during deglacial periods and underestimating NADW formation during glacial conditions.

Preferential dissolution of minerals with low ɛNd also alters the Nd isotopic composition of the residual sediment, particularly of the fine fraction, as shown by the ~6 ɛNd unit variation of silt in proglacial versus deglaciated coastal watersheds (Fig. 3). Therefore, reconstructions of ice sheet extent and circulation pathways that rely on the isotopic composition of the fine detrital fraction of marine sediments57,58,59,60 should consider variations in the extent of weathering of sediments. Our observed changes in particulate ɛNd with weathering extent are similar to or larger than the difference in ɛNd of two of the dominant terranes in south Greenland, the Nagssugtoqidian Mobile Belt (NMB) and the Archean Block (AB). Thus, the relative enrichment in radiogenic Nd produced by extensive weathering of the shield-derived sediments could be misinterpreted to reflect increasing contributions from younger terranes with more radiogenic Nd. Preferential mineral weathering also changes the Sr30,36 and Pb61 isotope ratios of the sedimentary suite, although in different proportions; thus, combining these three systems could further constrain changes in weathering extent versus sediment provenance.

Although chemical weathering rates depend on environmental conditions such as pH and redox potential, dissolution kinetics of Nd isotopes in terrestrial Precambrian mineral suites provide insights into the mechanisms that produce the ɛNd signature of the North Atlantic detrital and authigenic phases. Relatively rapid particle-seawater reactions, such as during suspended particle settling62, could result in lower seawater ɛNd, while longer interaction times, as in pore waters63, could generate more radiogenic Nd signatures. A greater focus on coupled authigenic-detrital data and mineralogy of detrital fractions could further enhance our understanding of Nd isotope systematics in the ocean.

Methods

Sampling

Stream water, bedload, and suspended sediment from SW Greenland were collected during the melt seasons in 2012, 2013, 2017, 2018, 2022, and 2023. For dissolved Nd isotopes, approximately 1–2 L of stream water were collected in acid-washed LDPE plastic containers by pumping and filtering through a 0.45 µm trace metal grade Geotech canister filter for all seasons except 2017 and 2018. In 2012 and 2013, samples for dissolved Nd isotopes were acidified to a pH <2 with ultrapure HNO3 and refrigerated until returned to the University of Florida (UF) laboratory for analyses. In 2022 and 2023, stream water samples were acidified to a pH of ∼3.5 with ultrapure HCl and pumped through a C18 cartridge previously cleaned and loaded with 2-ethylhexyl phosphate64 to extract REEs by complexation for transport to a UF laboratory for analysis. For analysis of dissolved Nd concentrations, 20 mL of 0.45 µm-filtered stream water were collected in acid-washed HDPE bottles, acidified to a pH <2 with ultrapure HNO3, and refrigerated for later analysis at UF. For bedload samples, approximately 500 g of sediment were collected with sterile plastic trowels or gloved hands, trying to minimize disturbing the sediment and maintain original grain size distribution. The samples were stored in Ziplock bags. In 2013, suspended sediments were collected by allowing 1 L of unfiltered water to settle, decanting ∼800 mL, and evaporating the remaining water. Suspended sediment samples from 2018, 2022, and 2023 were obtained by pumping unfiltered stream water through a 142 mm diameter, 0.45 µm cellulose nitrate filter enclosed in a Geotech tripod filter holder. The filter with the sediment was folded into quarters and stored in Ziplock bags.

Grain size separation

One bedload sample per watershed collected near the watershed outlet, was selected for grain size separation. Samples were freeze-dried and representative subsamples of ∼30 g for samples with abundant fine sediment and ∼250 g for those with minimal fine material were split using the coning and quartering technique. The subsamples were wet sieved through a 2 mm and then through a 63 µm sieve to obtain the sand fraction (63 µm – 2 mm). The <63 µm fraction was transferred to a 1000 mL graduated cylinder. Water was added until the water column height reached 10 cm. The sediment was then stirred for gravity separation of clay (<2 µm) and silt (2–63 µm) size fractions based on settling velocities calculated using Stoke’s Law. After settling for 7 h and 40 min, the liquid portion was carefully decanted into a beaker, ensuring minimal disturbance of the settled particles. Suspended sediments sampled across Sisimiut and AKR/WR streams were transferred to clean beakers by rinsing the cellulose filters. The bedload size fractions and suspended sediments were dried in an oven at 40 °C, weighed, and stored in clean vials for subsequent analyses. Throughout the procedure, only ultrapure water (18.2 MΩ.cm) and clean supplies were employed.

Nd concentration and Nd isotope analyses

Dissolved Nd concentrations were measured in a 1 mL aliquot of sample that was evaporated, re-dissolved in concentrated ultrapure HNO3 to oxidize organic matter, and then dried down and dissolved in 2.5% HNO3 spiked with Re-Rh. For dissolved Nd isotopic analyses, water samples collected in 2012 and 2013 were dried down in large beakers for several days. For water samples collected in 2022 and 2023 and processed through C18 cartridges, REEs were extracted by flushing the cartridges with 35 mL of 6 N HCl into Teflon beakers and evaporating the solution. The residues from both techniques were dissolved in concentrated ultrapure HNO3 to oxidize organic matter and dried. REE and then Nd aliquots were purified using column chromatography using TRU-Spec and Ln-Spec resins, respectively65.

For bulk bedload, grain size fractions, and suspended sediment samples, an ∼50 mg aliquot was weighed and transferred to Teflon beakers for dissolution and measurement of Nd isotope ratios and REE concentrations. Limited material for some of the suspended sediment and grain size fractions resulted in some samples as small as ∼1 mg. The solids were dissolved on a hot plate in a 5:1 mixture of HF-HNO3 in hex-cap Teflon vials. After dissolution, samples were dried and redissolved in concentrated ultrapure HNO3 and then 6 N ultrapure HCl to remove fluorides. For Nd concentration analyses, 4% of each sample was dissolved in 5% HNO3 spiked with Re-Rh. The remaining fraction was passed through Dowex 1X-8 ion exchange resin for separation of Pb66 followed by AG50W-X12 resin for REEs, and finally through Ln-Spec resin for Nd65.

REE concentrations were analyzed on an Element2 High Resolution Inductively Coupled Plasma Mass Spectrometer (HR-ICPMS) that was calibrated using USGS standards. Relative standard deviation (RSD) was better than 5% for concentrations greater than 0.016 ppb. Nd isotope ratios were measured on a Nu Multi Collector ICPMS (MC-ICPMS) using Time Resolved Analysis (TRA) software67. Based on repeat analyses of the JNdi-1 standard, the error on 143Nd/144Nd is ±0.000014 (2 SD), equivalent to ±0.3 ɛNd units.

Calculation of contribution to the Nd budget

Contribution percentage of each size fraction (\({\it{i}}\)) to the Nd budget of the bulk bedload was calculated from:

$${{{\mathrm{Contribution}}}}_{{i}}=\frac{{x}_{{i}}* {{{{\rm{C}}}}}_{{{\mathrm{Nd}}},i}* 100}{{\sum }_{j=1}^{n}({{{\it{x}}}}_{j}* {{{{\rm{C}}}}}_{{{\mathrm{Nd}}},j})}$$
(1)

where \({{{\it{x}}}}_{{{\it{i}}}}\) is the weight percentage of the fraction \({{\it{i}}}\); \({{{{\rm{C}}}}}_{{{\mathrm{Nd}}},i}\) is the Nd concentration of the fraction \({{\it{i}}}\); and \({{\it{n}}}\) represents the number of fraction constituents in the bedload.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.