Observations in the LESS were conducted on board the Russian R/V Akademik Tryoshnikov from September to October 2021 (Fig. a). Low-density polyethylene (LDPE) bottles and buckets (Thermo Fisher Scientific, USA), polyethylene bags (GL Sciences, Japan), AcroPak capsules with Supor membrane filters (0.8/0.2 µm pore size, Pall, USA), and a Tygon tube (Masterflex, Germany) used for sampling trace metals were thoroughly acid-cleaned in a class-100 clean-air laboratory. Seawater was collected from a depth of approximately 10 m from the side of the ship using a peristaltic pump (Geopump, Geotech Environmental Equipment, USA) and a Tygon tube. Filtered and unfiltered Fe and Mn samples were obtained to assess their labile particulate fractions. Samples of the dissolved fractions of Fe and Mn (dFe and dMn) were collected in LDPE bottles after filtration through AcroPak filters connected to Tygon tubes. Samples for total acid dissolvable Fe and Mn (TdFe and TdMn) were collected in LDPE bottles without filtration. The pH of the Fe and Mn samples was adjusted to < 1.8 by adding ultrapure-grade 6 M hydrochloric acid (Tamapure AA-100, Tama Chemicals, Japan), and the samples were stored for a year before the analysis. Therefore, the concentration differences between the unfiltered (i.e., TdFe and TdMn) and filtered (i.e., dFe and dMn) samples were attributed to the acid-labile particulate fraction. Samples for nutrient analysis were collected in acrylic vials after filtration. Samples for dissolved organic matter (DOM) were collected in precombusted (450 °C for 5 h) glass bottles after filtration through precombusted glass fiber filters (0.7 µm nominal pore size, Whatman GF/F, UK). The samples for the stable isotope of oxygen (δ18O) in the seawater were collected in glass bottles with rubber inserts in the caps. The samples for nutrients and DOM were frozen immediately after collection at -20 °C and were shipped back to the onshore laboratory.

The water properties at depths of 0–30 m were measured using a portable conductivity–temperature–depth (CTD) sensor (RINKO 102, JFE Advantech, Japan) and are shown in a temperature–salinity diagram (Fig. S1 in the Supplement). Full-depth temperature and salinity profiles were obtained using a Seabird SBE911plus CTD system (Figs. S2 and S3).

Sea ice observations were made at three ice stations on 1 October (St. Ice 1), 2 October (St. Ice 2), and 10 October (St. Ice 3) 2021 (Fig. a and b). Water sampling under sea ice was performed to analyze Fe, Mn, nutrients, DOM, and δ18O at depths of 1, 5, and 10 m using the pump system described in Sect. . Sea ice cores were collected using an ice corer (Mark II Coring System, Kovacs Enterprise, USA) and sectioned into five subsamples using a titanium flat-headed screwdriver. The subsamples were cleaned at the site by removing more than 2 cm of their outer layers using acid-cleaned ceramic knives, following previously reported methods (Evans and Nishioka, 2019; Kanna et al., 2014). The cleaned ice samples were transferred to LDPE buckets. Snow samples were collected in polyethylene bags by using an acid-cleaned polycarbonate scoop. The snow and cleaned ice samples were melted inside a class-100 clean-air bench in the laboratory. The meltwaters for Fe, Mn, nutrients, DOM, and δ18O analyses were subsampled and processed as described in Sect. . The salinity of the meltwater samples was measured using a portable salinity sensor (Multi 3510; Xylem, USA).

The acidified water samples for Fe and Mn analyses were preconcentrated using a manual solid-phase extraction system equipped with a Nobias Chelate-PA1 resin column (Hitachi High Technologies, Japan) (Sohrin et al., 2008; Kondo et al., 2016; Kanna et al., 2022). Ultrapure-grade nitric acid, acetic acid, and ammonium solution (Tamapure AA-100, Tama Chemicals, Japan) were used for preconcentration and extraction. The water samples were adjusted to pH 6.0 ± 0.1 by adding a 3.6 M ammonium acetate buffer solution prepared from the acetic acid and ammonium solution. Fe and Mn concentrated on the resin were eluted with 2 M nitric acid and analyzed using high-resolution inductively coupled plasma mass spectrometry (ELEMENT XR, Thermo Fisher Scientific, USA). Procedure blanks of Fe and Mn were evaluated using ultrapure water following the preconcentration procedures that showed 0.13 ± 0.04 nmolkg-1 (n=84) for Fe and 0.002 ± 0.005 nmolkg-1 (n=84) for Mn, respectively. Certified reference materials for trace metals, NASS-7 and CASS-6 (National Research Council of Canada), were used to validate the preconcentration procedures. The analytical values were within the error range of the certified reference materials (Table ).

Dissolved organic carbon (DOC) was analyzed using a total organic carbon analyzer (TOC-L, Shimadzu, Japan). Absorbance spectra for chromophoric dissolved organic matter (CDOM) were analyzed between 200 and 800 nm at 1 nm intervals using a dual-beam spectrophotometer (UV-1800, Shimadzu, Japan) with a 1 cm quartz-windowed cell. The sample spectra were corrected using ultrapure water spectra and converted to Napier absorption coefficients at wavelength (a (λ), m-1) (Green and Blough, 1994).

The excitation–emission matrix fluorescence spectra of the CDOM were analyzed using a fluorescence spectrophotometer (FP-8500, JASCO, Japan). The spectra were obtained at excitation wavelengths ranging from 250 to 500 nm and emission wavelengths ranging from 280 to 600 nm. The fluorescence intensity was corrected to the area under the Raman peak of ultrapure water (excitation wavelength = 350 nm) and calibrated to Raman units (RUs) using the methods outlined by Lawaetz and Stedmon (2009) and Tanaka et al. (2016).

The δ18O value of the water samples was determined using an isotope water analyzer (Picarro L2120-i, Picarro, USA) with an analytical precision of ± 0.3 ‰. Macronutrient concentrations were determined using an autoanalyzer (QuAAtro, BL TEC, Japan) with a continuous-flow system. The measurements were calibrated using seawater as a reference (KANSO Technos, Japan).

The CDOM absorption coefficient at 350 nm (a350) was used to indicate terrestrial humic substances. In addition, parallel factor analysis (PARAFAC) was applied to statistically decompose the excitation–emission matrix fluorescence spectra into their components. PARAFAC was performed in MATLAB (MathWorks, Natick, MA, USA) using the DOMFluor toolbox (Stedmon and Bro, 2008). The dataset used in this study comprises 42 samples of seawater, snow, and sea ice. A brine sample (n=1) was assessed as an outlier in the PARAFAC model and was not used in this study. The wavelengths for PARAFAC were obtained at excitation wavelengths ranging from 250 to 500 nm and emission wavelengths ranging from 280 to 535 nm. The correct number of components was primarily determined using split-half analysis and random initialization (Stedmon and Bro, 2008). The three-component model was validated using the PARAFAC model (Fig. S4). Components 1 and 2 exhibit fluorescence peaks in the visible region, which is defined as visible fluorescence. Component 1 peaks at an emission wavelength of 410 nm and is traditionally categorized as a humic-like fluorophore of marine origin (Coble, 1996). Component 2 peaks at an emission wavelength of 470 nm and is traditionally categorized as a humic-like fluorophore of terrestrial origin (Coble, 1996), although it has been reported that marine microbes produce this type of fluorophore (Goto et al., 2020). In this study, component 1 was combined with component 2 and interpreted as all humic-like fluorophores because their fluorescence intensities were not very different among the sampling locations. Component 3 exhibits fluorescence peaks in the ultraviolet A (UVA) region, which is defined as UVA fluorescence. This fluorophore is traditionally categorized as a protein-like fluorophore (Coble, 1996).

To quantify the relative contribution of freshwater to changes in the surface water properties of the LESS, we assumed a mixture of four components: meteoric water, sea ice meltwater, Pacific Water, and Atlantic Water. Fractions of mass, salinity, and δ18O together with the nitrate : phosphate ratio balance equations are described using the water properties of these four endmembers (Bauch et al., 2011; Charette et al., 2020; Gerringa et al., 2021; Newton et al., 2013): 1fmw+fsim+fPacific+fAtlantic=1,2fmw⋅Smw+fsim⋅Ssim+fPacific⋅SPacific+fAtlantic⋅SAtlantic=S,3fmw⋅δmw+fsim⋅δsim+fPacific⋅δPacific+fAtlantic⋅δAtlantic=δ,4fmw⋅Pmw+fsim⋅Psim+fPacific⋅PPacific+fAtlantic⋅PAtlantic=P, where f, S, δ, and P are the fractions of mass, salinity, δ18O, and phosphate concentration based on the nitrate : phosphate ratio and the suffixes mw, sim, Pacific, and Atlantic indicate meteoric water, sea ice meltwater, Pacific Water, and Atlantic Water, respectively (Table ). fmw includes river runoff and local precipitation. The measured values of nitrate + nitrite (N) in each sample were used to compute the individual phosphate endmembers for the Pacific and Atlantic fractions of the Atlantic Water Line (P=0.0596×N+0.1139; Bauch et al., 2011) and Pacific Water Line (P=0.0653×N+0.94; Jones et al., 2008) for each sample. Notably, this calculation produced a slightly negative fPacific value owing to inaccuracies in the endmembers and measurements (Bauch et al., 2011). Indeed, our calculation showed negative fPacific values at the four stations, with an average of -2 %. However, this error remains relatively small within the Atlantic regime and is still within the uncertainty (∼ 10 % for marine waters) of the method (Yamamoto-Kawai et al., 2008).

Water properties in the East Siberian Sea and Chukchi Abyssal Plain differed significantly from those of the Makarov and Amundsen basins in the late summer of 2021. The surface waters in the East Siberian Sea and Chukchi Abyssal Plain were relatively cold (temperature < 0 °C) and fresh (salinity < 30) compared with those in the Makarov and Amundsen basins (Figs. b, c and S2). The waters in the East Siberian Sea and Chukchi Abyssal Plain were characterized by low δ18O values of less than -2.5 ‰ (Fig. d). The differences between the regions were likely due to the magnitude of the mixing of Pacific-sourced water with Atlantic-sourced water, river runoff, and the melting of sea ice. Pacific-sourced water enters the Bering Strait, passes through the Chukchi Sea, and penetrates the East Siberian Sea (Fig. a). Atlantic-sourced water enters the Amundsen Basin and flows along a continental slope (Fig. a). The mixed layers in the Atlantic and Pacific sectors have different geochemical and physical characteristics. In the temperature–salinity diagram, the properties of the upper water layers (< 30 m) of the East Siberian Sea and Chukchi Abyssal Plain (Sts. 71 and 89 and Ice 1; Fig. b) were similar to those of the Surface Polar Mixed Water that originated in the Pacific sector of the Arctic Ocean, which is characterized by temperatures < 0 and salinity < 31 (Figs.  and S1). In contrast, the upper water layers found in the Makarov and Amundsen basins (Sts. 1, 11, and 31 and Ice 3; Fig. b) are not derived from Surface Polar Mixed Water but are a product of a mixture of warm, saline Atlantic Water (temperature > 0 and salinity > 34.8) and freshwater (Figs.  and S1).

The surface water in the East Siberian Sea and Chukchi Abyssal Plain was generally enriched in silicate and Mn but depleted in Fe relative to the Makarov and Amundsen basins (Fig. a–e). The concentrations of silicate and TdMn reached 8.9 µmolL-1 and 83.5 nmolkg-1, respectively, and the concentration of TdFe was as low as 0.9 nmolkg-1 in the East Siberian Sea and Chukchi Abyssal Plain. The TdFe concentration tended to be high along the continental slope (Fig. e), with a maximum value of 79.2 nmolkg-1. Differences between the Fe and Mn distributions were also observed at the two shallow (∼ 40 m bottom depth) southernmost stations on the shelf of the East Siberian Sea (Sts. 86 and 89; Fig. ). The TdFe concentrations were relatively low (1.58–1.62 nmolkg-1) at the stations where the notably high TdMn values were detected (70.7–83.5 nmolkg-1). dFe and dMn, which did not include particles, exhibited spatial variations similar to those of TdFe and TdMn. The concentration differences between the unfiltered (i.e., TdFe and TdMn) and filtered (i.e., dFe and dMn) samples are attributed to the acid-labile particulate fractions of these metals. The acid-labile particulate fractions of Fe accounted for 48 %–61 % of the TdFe but only 6 %–18 % of the TdMn at the stations on the shelf. These results indicate that Fe was relatively abundant in the particulate phase and Mn in the dissolved phase on the shelf. Various DOC concentrations were observed at the different sampling locations (Fig. f) and exceeded 100 µmolL-1 at most of the stations. The detailed DOC distributions are discussed in Sect.  along with the optical information of the CDOM.

The mass fractions of the Pacific Water (fPacific), Atlantic Water (fAtlantic), sea ice meltwater (fsim), and meteoric water (fmw) components computed by solving Eqs. ()–() are shown in Fig. a–d. fPacific was greater than 50 % in the surface waters of the East Siberian Sea and Chukchi Abyssal Plain and less than 20 % in the Makarov Basin (Fig. a). Small values of fPacific (up to 2 %) were observed in the Amundsen Basin. At this location, however, the fPacific signal does not originate from Pacific-sourced water but from water modifications by denitrification within the sediment over the Laptev Shelf (Bauch et al., 2011). A negative or positive value of fsim indicates the addition of brine or meltwater to the surface layer (Fig. c). The fsim values in the surface water were as high as 5.8 % over the East Siberian Sea and 3.4 % over the Amundsen Basin (Fig. c), where the melting of sea ice was likely predominant. In contrast, a negative value of fsim in the Makarov Basin suggests that sea ice formation is dominant, which agrees with previous studies (Bauch et al., 2011; Yamamoto-Kawai et al., 2005). The fmw in the surface water increased by approximately 20 % from west to east (Fig. d). The long-term trend of fmw has increased in the Pacific sector of the Arctic Ocean since 1981 (Polyakov et al., 2020), which may be due to the influence of Siberian river runoff or increased freshwater flux through the Bering Strait.

During the NABOS expedition in the late summer of 2021, snow, sea ice, and under-ice water were collected from three ice stations (Fig. ). Detailed observations of the ice cores and snow pits at these stations can be found in the NABOS 2021 Cruise Report (https://uaf-iarc.org/nabos/cruise-reports/, last access: 2 November 2024). Thin-section analysis of a single ice core collected at St. Ice 2 revealed that the sea ice was composed of approximately 9 % granular ice, 37 % columnar ice, and 54 % mixed ice (Fig. S5). Salinity and δ18O values in the sea ice were similar between the stations, ranging from 0 to 3.8 and from -3.6 ‰ to 0 ‰, respectively (Fig. a and b). Generally, Fe and Mn concentrations in the sea ice were lower than those in the under-ice water, except for the ice section (0.2 m from the bottom) collected at St. Ice 3 (Fig. c–f). The dFe concentrations in the sea ice gradually decreased from 3.2 to 1.7 nmolkg-1 with the ice core depth, while the dMn concentrations increased from 2.4 to 5.2 nmolkg-1. The TdFe and TdMn concentrations exhibited vertical variations, similar to those of the dissolved fractions. The TdFe concentration tended to be higher in the snow samples than in the sea ice and under-ice water; however, this did not apply to TdMn.

We calculated the inventories of Fe and Mn from the cumulative metal loads of the 110 cm ice cores collected at St. Ice 1 (Table ). The metal inventory of the ice core was comparable to that obtained from the Canada Basin (Evans and Nishioka, 2019), except for TdFe, which had a lower inventory value. Moreover, the reported particulate Fe inventory in the ice core from the western and central Arctic Ocean (Bolt et al., 2020) exceeded the TdFe inventory in this study. Such discrepancies are likely due to greater heterogeneity in the distribution of particulate Fe loads in Arctic pack ice (Bolt et al., 2020). The sediment loadings of the collected ice samples were low based on visual observations, such that there was no evidence of ice-rafted sediment adding Fe and Mn to the surface waters, although their importance has been reported (Hölemann et al., 1999a, b; Measures, 1999; Rogalla et al., 2022; Tovar-Sánchez et al., 2010).

The DOC concentration varied in the snow (24.5–105 µmolL-1) and sea ice (32.1–147 µmolL-1) samples, while the concentrations were relatively uniform vertically in the under-ice water (84–110 µmolL-1) (Fig. g). The optical properties of the CDOM showed that the intensities of the visible fluorescence were lower in the snow and sea ice than in the under-ice water (Fig. h). The intensity of the UVA fluorescence was relatively high in the sea ice at St. Ice 2 but not at St. Ice 1 (Fig. i). Macronutrients were generally depleted in the sea ice, except for nitrate + nitrite at St. Ice 2, which showed an enrichment of 1.3 µmolL-1 compared to the under-ice water (< 0.25 µmolL-1) (Fig. j–l). The enrichment of nitrate + nitrite was also observed in the snow samples (1.2 µmolL-1).

Variations in the salinity and δ18O values in the surface water of the LESS continental margins are attributed to local precipitation or river runoff and melting or formation of sea ice. Figure  shows the salinity–δ18O diagram based on the samples of the surface water, snow, sea ice meltwater, and previously reported freshwater sources (Evans and Nishioka, 2019; Marsay et al., 2018; Peterson et al., 2016). The salinity and δ18O values in most of the surface water samples deviated from those of the meteoric endmember, including the snow meltwater and the Siberian rivers. However, the salinity and δ18O values in the under-ice water at St. Ice 2 rather deviated toward the sea ice endmember, which resulted from substantial input of sea ice meltwater into the station.

The dMn concentration in the surface water gradually increased with decreasing salinity (Fig. b). The dFe concentration also showed a similar increasing trend in salinity ranges from 31 to 34 (Fig. a); however, the dFe concentration in less saline waters (salinity < 31) was independent of the salinity variation. Previous studies have shown notably high dFe and dMn concentrations in the Siberian rivers of 1751 ± 1218 and 208 ± 279 nmolL-1, respectively (Peterson et al., 2016), such that river waters must have supplied substantial amounts of Fe and Mn to the surface of the LESS. In contrast, dFe and dMn concentrations in the sea ice endmember were relatively low (2.0 ± 1.3 nmolkg-1 for dFe and 5.3 ± 3.3 nmolkg-1 for dMn). The enriched dMn in the under-ice water at St. Ice 2 (13.5 nmolkg-1) was likely not derived from the input of sea ice meltwater because of the low level of dMn in the sea ice endmember, suggesting the existence of extra Mn sources in the surface water. In the subsequent section, we discuss the factors controlling the distribution of Fe and Mn on the surface of the LESS.

Based on the water mass analysis described in Sect. , we classified the surface water as Surface Polar Mixed Water (PMW) and Surface Atlantic Mixed Water (AMW) (Fig. a and b; the dotted circles border the area). The PMW contains a large fraction of Pacific-sourced water (fPacific > 50 %) and is mostly found in the East Siberian Sea and Chukchi Abyssal Plain, whereas the AMW is found in the Makarov and Amundsen basins. The PMW is generally characterized by high-silicate, low-Fe, and high-Mn concentrations relative to the AMW (Fig. a–e). The high silicate content of the PMW likely resulted from nutrient-rich Pacific Water entering the shallow Bering Strait (Chen et al., 2018; Jensen et al., 2020; Nishino et al., 2013). The maximum surface silicate concentration along the continental margin of the western Makarov Basin, which is a typical signature of runoff from the Lena River in the region (Alling et al., 2010; Anderson et al., 2017), was not observed in this study (Fig. a).

A significant correlation between dFe and dMn has been observed in the deeper waters (> 3000 m) of the Amundsen and Makarov basins because scavenging removal is the dominant process in deep-water masses (Klunder et al., 2012). In the LESS surface water, Fe was not correlated with Mn in the unfiltered and filtered fractions (Fig. S6). Additional factors, such as external inputs, influence the distribution of dFe and dMn in surface waters, leading to the disappearance of the Fe–Mn relationship. Moreover, the enrichment of dMn compared to that of dFe (Fig. S6a) was observed in all the sampled surface waters, suggesting the importance of the input fluxes of dMn or the preferential scavenging of dFe relative to dMn. Fe and Mn are redox-active metals that share common sources in surface water, such as sediments, dust deposition, and freshwater inputs. The dust deposition is considered to be of minor importance in the LESS, given the relatively low concentrations of TdFe (∼ 19.9 nmolkg-1) and TdMn (∼ 1.1 nmolkg-1) observed in the snow samples in this study (Fig. d, f). In general, the summertime atmospheric deposition fluxes of Fe and Mn into the Arctic Ocean are reportedly low (Kadko et al., 2016; Marsay et al., 2018), especially in comparison to Arctic rivers and coastal erosion and diagenetic fluxes from shelf sediments (Charette et al., 2020; Jensen et al., 2021; Kadko et al., 2018). In the following discussion, we evaluate the potential sources (sedimentary input, river runoff, and sea ice formation and melt) of these metals in the PMW and AMW.

In the surface LESS, various concentrations of DOC were observed, with the highest value of 146 µmolL-1 (Fig. f). The mean value of DOC (108 ± 19 µmolL-1) was similar to the value previously observed in the Laptev and East Siberian seas (Hölemann et al., 2021) but tended to be high compared to the Chukchi Sea (Chen et al., 2018; Jung et al., 2021; Tanaka et al., 2016). In addition to DOM release from productive shelf sediments (Cooper et al., 2005), regional DOM inputs from East Siberian rivers (Stedmon et al., 2011) and inputs of fresh-plankton-derived DOM (Davis and Benner, 2007) result in surface waters of the LESS with a high DOC signal. This study investigated the optical properties of CDOM, which are useful for understanding the composition and origin of the DOM in the studied region.

The DOC concentrations in the AMW and PMW were positively correlated with the absorption coefficient a350 and the intensities of visible fluorescence (Fig. a, b). This correlation implies that the factors controlling the DOC concentrations and the abundance of humic-like CDOM are similar in these surface waters. The intensities of visible fluorescence in the AMW and PMW were strongly correlated with fmw (Fig. c), suggesting the importance of riverine humic-like CDOM sources in surface waters. Interestingly, the linear fits of the visible fluorescence–fmw relationships were different between the AMW and PMW (Fig. c). This result indicates that the riverine endmembers of humic-like CDOM were different between the AMW and PMW; i.e., the Lena River was the riverine endmember of the AMW, whereas the Indigirka and/or Kolyma rivers were the endmembers of the PMW. Microbial processing of DOM can also generate visible fluorescence (Nelson et al., 2004; Rochelle-Newall and Fisher, 2002; Yamashita and Tanoue, 2008). The low-N∗ PMW must have received microbially mediated visible fluorescence from the benthic remineralization of sinking organic matter on the productive Chukchi Sea shelf (Hioki et al., 2014; Yamashita et al., 2019).

In addition to visible fluorescence, the intensity of UVA fluorescence in the PMW was positively correlated with DOC concentration; however, this does not apply to the AMW (Fig. c). In the Arctic Ocean, UVA fluorescence is regarded as a phytoplankton-derived component (Brogi et al., 2019), including its direct release from phytoplankton and its release during grazing by zooplankton. We integrated chlorophyll fluorescence into the water column (0–10 m depth) of the PMW and compared it with the UVA fluorescence intensity (Fig. S8). The linear correlation between chlorophyll fluorescence and UVA fluorescence in the PMW (r=0.80; p < 0.01) indicates that the local production (and degradation) of UVA fluorescence is an important process. In addition, the intensity of UVA fluorescence in the PMW and AMW increased with increasing fmw (Fig. d). This result can be interpreted as the rapid transport of freshly produced labile CDOM by river water to the surface of the LESS.

The relationship between fmw and trace metal concentrations revealed that the riverine source of dFe behaved more conservatively than that of dMn (Fig. a, b), which could be explained by the existence of organic ligands complexed with Fe. Previous studies on the Arctic Ocean reported that the CDOM pool contains Fe-binding organic ligands in the form of humic substances (Laglera et al., 2007, 2011; Laglera and van den Berg, 2009; Slagter et al., 2017; Williford et al., 2021). Complexes of Fe-humic substances account for approximately 80 % of the dFe concentrations in the Arctic Ocean, and the concentrations of these complexes are highly correlated with the CDOM and dFe concentrations (Laglera et al., 2019). Our results show significant correlations between dFe and visible fluorescence in both the AMW and PMW (Fig. a). The input of humic substances from Siberian rivers into these waters was evident from the elevated intensities of visible fluorescence with fmw (Fig. c). Thus, humic substances strongly affect the dFe concentration in seawater through complexation with Fe, which stabilizes Fe in the dissolved phase. A difference in the linear fits of dFe–visible fluorescence relationships between the AMW and PMW (Fig. a) may be explained by the contribution of excess humic ligands from rivers to the PMW. In addition to Fe, Mn is known to form organic complexes with the degradation products of organic matter, such as humic materials (e.g., Oldham and Tebo, 2017) and biogenic siderophores (e.g., Parker et al., 2004). However, complexation of organic ligands with Mn in the Arctic Ocean is poorly understood. None of the AMW or PMW samples exhibited a significant correlation between dMn and visible fluorescence (Fig. b). Unlike dFe, dMn is not stabilized by humic-type organic complexes in surface waters.

To investigate the distribution patterns of dFe and dMn in the surface waters, we combined our dataset with data available for the Arctic Ocean (Cid et al., 2011, 2012; GEOTRACES Intermediate Data Product Group, 2023; Gerringa et al., 2021; Hölemann et al., 2005; Jensen et al., 2020; Klunder et al., 2012; Kondo et al., 2016; Middag et al., 2011; Rijkenberg et al., 2018; Savenko and Pokrovsky, 2019). The dFe concentrations are relatively low (up to 2 nmolkg-1) in the Atlantic sector of the Arctic Ocean (Fig. a). At the surfaces of the Nansen Basin and Barents Sea, Fe is expected to be the first nutrient to be depleted by primary producers (Rijkenberg et al., 2018), and phytoplankton consumption could be an important sink for Fe. On the LESS continental margins, however, surface dFe concentrations are significant (> 5 nmolkg-1) and even persist in the late summer of 2021 (Fig. a). Figure a shows a boxplot of the surface dFe concentration combined with our dataset, with the available data reported for each region in the Arctic Ocean (Fig. a and c). All the water samples in Fig.  had salinities greater than 25. The result shows that the median value was highest in the Bering Sea (4.7 nmolkg-1), followed by the Chukchi Sea (3.1 nmolkg-1) and the LESS (2.8 nmolkg-1). We deduce that sedimentary Fe originating from the reductive Bering Sea shelf was gradually removed from the Chukchi Sea after entering it from the Bering Strait (Jensen et al., 2020), and it then penetrated the East Siberian Sea. The dFe concentrations also increased in the estuaries of the Lena, Yenisei, and Mackenzie rivers (Fig. a). Other studies determined that the dFe concentration in the estuary water (salinity < 25) of the Lena, Yenisei, and Mackenzie rivers was as high as 9000 nmolkg-1. Fe-binding organic ligands in the form of humic substances originating from the Lena River strongly affect the dFe concentration, as discussed in the previous section. The natural humic substance Fe-binding ligands of the Arctic Ocean surface belong to a group of strong ligands ubiquitous in surface ocean waters (Laglera et al., 2019). Strongly complexed Fe may be less biologically available to the phytoplankton community than weakly complexed Fe released from grazing and bacterial remineralization of organic matter (Gledhill and Buck, 2012). The river-influenced water from the LESS continental margins is the source water of the Transpolar Drift, which enriches the dFe in the central Arctic Ocean (Fig. a) (Charette et al., 2020; Gerringa et al., 2021; Klunder et al., 2012).

The high dMn value found in the central Arctic Ocean (Fig. b) was also related to the presence of the Transpolar Drift (Charette et al., 2020; Gerringa et al., 2021). In addition to riverine inputs, sediment–water column exchange over shelves leads to relatively dMn-rich water in the Pacific sector of the Arctic Ocean. The dMn concentrations increased toward the broad shelves of the East Siberian Sea (∼ 68 nmolkg-1), Chukchi Sea (∼ 45 nmolkg-1), and Bering Sea (∼ 103 nmolkg-1), whereas the concentration was relatively low (∼ 8 nmolkg-1) over a narrow shelf of the Beaufort Sea (Fig. b). A boxplot of the surface dMn concentration in Fig. b shows that the difference in median values is relatively small in the LESS (13.0 nmolkg-1), Chukchi Sea (12.4 nmolkg-1), and Bering Sea (12.2 nmolkg-1). As discussed previously, the shelf sediment–water exchange processes over the Chukchi Sea largely influence the Fe and Mn distributions in the East Siberian Sea. Vieira et al. (2018) provided the first estimate of the benthic flux of the radium isotope (Ra228) in the Chukchi Sea as a tracer of benthic trace metal inputs, which was among the highest rates reported globally. The low-N∗ water spread over regions where nitrate was already depleted relative to phosphate, mainly because of the oxidation of organic matter by bacteria in the reductive shelf sediment (Fig. c). The East Siberian Sea and Chukchi Sea are likely hotspots of sediment-sourced dMn via the reductive dissolution of Mn oxide in the sediment. A multistep removal process of dMn has been suggested in the Arctic (Jensen et al., 2020): dMn is removed rapidly in the particulate phase within 150 km of the shelf break, but some dMn remains preserved throughout the next 1000 km away from the shelf. The dMn from the LESS continental margins appears to be exported effectively by the Transpolar Drift to the central Arctic Ocean (Fig. b), although stabilization by organic complexation is unlikely for Mn during offshore transport.