A hydrographic survey was conducted at a total of 29 stations in the western Arctic Ocean during the ARA10B cruise (August 5−24, 2019) aboard the Korean icebreaker IBR/V Araon (Figure 1). The study area was divided into three regions based on its geographical location, namely the Chukchi Sea (CS; stations 1–13), Chukchi Borderland (CBL; stations 14–26 and 34), and East Siberian Sea (ESS; stations 27–33). In order to explore the water column characteristics in different regions, we partitioned the sampling stations into two transects, namely transect 1 across the CS and CBL and transect 2 across the ESS and CBL.

Seawater sampling was conducted using 10 L Niskin bottles on the surface and at discrete depths chosen based on conductivity-temperature-depth (CTD; SBE 911 plus, SeaBird Electronics Inc., Bellevue, WA, USA) profiles. Seawater samples for DOM optical properties, DOC, and oxygen isotope ratios (δ¹⁸O) were gravity filtered from the Niskin bottles using an inline pre-combusted (at 550°C for 6 h) Whatman GF/F filter held in an acid-cleaned (0.1 M HCl) polycarbonate 47-mm filter holder (PP-47, ADVANTEC), as described in detail by Jung et al. (2021b). In each case, the filter holder was attached directly to the Niskin bottle spigot. For DOM optical properties and DOC measurements, the filtrates were divided into pre-combusted (at 550°C for 6 h) 20 mL glass ampoules using a sterilized serological pipette. For δ ¹⁸O measurement, the filtrate was distributed into an acid-cleaned 20 ml glass vial. The ampoules for DOM optical properties and DOC were sealed with a torch, quick-frozen, and stored under dark conditions at –24 °C for approximately two months until the analysis in our land laboratory. The vials for δ¹⁸O were sealed with parafilm and stored at 4 °C for approximately two months until the analysis in our land laboratory. For inorganic nutrient measurement, seawater was drawn into 50 mL conical tubes and immediately stored in a refrigerator at 4 °C until chemical analysis was conducted onboard within three days. Seawater samples (300–500 mL) for Chl-a analysis were filtered through Whatman GF/F filters. The filters were then immersed in 90% acetone for 24 hr to extract the Chl-a. The Chl-a concentration was immediately measured onboard after the extraction process. Seawater samples for heterotrophic bacterial abundance measurement were transferred into 2 mL cryovials, fixed with a final concentration of 1% (v/v) paraformaldehyde for 15 min, and stored at −80°C. The samples were analyzed onboard within a few weeks after collection.

When considering the filtration of FDOM samples, the choice of filter pore size is an important factor that can influence the properties of FDOM. While some studies have recommended using 0.2 μm pore size filters to prevent the passage of microorganisms and colloids that are not considered true DOM (Castillo and Coble, 2000; Moran et al., 2000; Yamashita et al., 2008; Romera-Castillo et al., 2010; Yao et al., 2011; Jørgensen et al., 2014; Gonçalves-Araujo et al., 2015; Logvinova et al., 2015; Logvinova et al., 2016; Mendoza et al., 2017), there is also a widespread use of GF/F filters with a pore size of 0.7 μm in previous studies (Stedmon et al., 2003; Yamashita and Tanoue, 2003; Stedmon and Markager, 2005a; Stedmon and Markager, 2005b; Stedmon et al., 2011; Dainard and Guéguen, 2013; Yamashita et al., 2013a; Gao and Guéguen, 2017; Yamashita et al., 2017; Chen et al., 2018; Brogi et al., 2019; Dainard et al., 2019; Kim et al., 2020; Drozdova et al., 2022) due to their high flow rate, high loading capacity, and the ease of the cleaning procedure (Spencer and Coble, 2014). Additionally, Nayar and Chou (2003) found that pre-combusted GF/F filters had an average pore size of 0.3 μm. Hence, taking into account the availability and performance characteristics of the pre-combusted GF/F filters, the use of these filters for FDOM filtration is expected to yield comparable results to 0.2 μm filters.

Absorbance spectra of filtered seawater samples were measured using a spectrophotometer (UV-2600, Shimadzu Inc., Japan) at 1 nm intervals between 200 and 800 nm. All sample spectra were corrected using Milli-Q water measured daily. The measured absorbance was converted to absorption coefficient, according to the following equation:

where a is the absorption coefficient (m^–1) at wavelength λ, A is the absorbance at wavelength λ, and L is the path length (m) (Green and Blough, 1994).

The spectral slope coefficient between 275 and 295 nm (S_275–295, nm^–1) was calculated using an exponential function, as described by Green and Blough (1994):

where a(λ) and a(λ₀) are absorption coefficients at wavelength λ and λ₀ (λ > λ₀).

S_275–295 has been widely applied in studies of aquatic environments (Fichot and Benner, 2012; Granskog, 2012; Granskog et al., 2012; Dainard and Guéguen, 2013; Helms et al., 2013; Yamashita et al., 2013b; Lee et al., 2016; Shen et al., 2016a; Johnston et al., 2018; Hirawake et al., 2021), because it can provide insight into the source, DOM molecular weight, and photochemical alterations (Helms et al., 2008; Helms et al., 2013).

Fluorescence EEMs were obtained using a fluorescence spectrophotometer (F-7100, Hitachi Inc., Japan) at excitation/emission (Ex/Em) wavelengths of 250−500/280−550 nm. The Ex and Em scans were set at 5 nm and 1 nm steps, respectively. Milli-Q water was used as a blank and its EEM spectrum was subtracted from the EEM spectra of the sample being analyzed. The fluorescence intensities were normalized to the area under the Milli-Q water Raman peak and reported in Raman units (R.U.) (Stedmon et al., 2003). Since absorbance values were always lower than 0.3 at 254 nm for all samples in this study, both reabsorption and inner filter effects were minimized (Burdige et al., 2004). The EEMs were characterized by PARAFAC modeling in MATLAB (Mathworks, USA) using the DOMFluor toolbox (Stedmon and Bro, 2008). The number of components was determined by a split-half validation and the percentage of explained variance (98.5%).

DOC concentration was measured by high-temperature combustion using a TOC analyzer (TOC-L, Shimadzu Inc., Japan). The concentration of DOC in Milli-Q water was measured daily and used as a blank. DOC concentration in consensus reference materials (CRM, 42−45 µM C for DOC, deep Florida Strait water obtained from the University of Miami, USA) were measured at every sixth analysis to check the reliability of measurements. The instrument ran 3–4 injections per sample, and analytical errors were within 5% for DOC (Chen et al., 2018; Jung et al., 2021b).

Samples for δ¹⁸O measurement were analyzed by equilibration with carbon oxide. The δ¹⁸O was measured using a stable isotope ratio mass spectrometer (Isoprime, Micromass, Manchester, UK) in Korea Basic Science Institute. Data are reported with respect to Vienna-Standard Mean Ocean Water (V-SMOW) standard with the δ¹⁸O notation, where δ¹⁸O = [(¹⁸O/¹⁶O_sample/¹⁸O/¹⁶O_V-SMOW)–1] × 1000. Precision of the analysis, based on duplicate measurements of an internal standard, was determined to be < 0.1‰ (Jung et al., 2021b).

δ¹⁸O and S were used to estimate the fractions of seawater (f_seawater), sea ice meltwater (f_{sea ice melt}), and river water (f_river), by assuming that the observed δ¹⁸O and S in the seawater samples have resulted from a mixture of river water, sea ice meltwater, and seawater. The respective fractions (i.e., f_seawater, f_{sea ice melt}, and f_river) were calculated using the following mass balance equations:

where f and S are the fraction and salinity, respectively. δ¹⁸O_obs and S_obs refer to the observed values from each seawater sample. We chose the three end-member values for the calculation of f_river, f_{sea ice melt}, and f_seawater (Table 1), as suggested in previous studies (Mathis et al., 2007; Cooper et al., 2008; Yamamoto-Kawai et al., 2008).

Inorganic nutrients, including nitrite + nitrate (NO₂^– + NO₃^–), and phosphate (PO₄^3–), were measured onboard using a four-channel auto-analyzer (QuAAtro, Seal Analytical, Germany) following the Joint Global Ocean Flux Study (JGOFS) protocols (Gordon et al., 1993). The reference material for nutrients in seawater (Lot No. “BV”, KANSO Technos Co., Ltd., Osaka, Japan) was used in conjunction with standards for each batch of runs to ensure precise and consistent measurements. The accuracies of the measured concentrations (reported as relative standard deviation) of NO₂^– + NO₃^– and PO₄^3– were ± 1.0% at 35.33 μmol kg^–1 and ± 0.89% at 2.514 μmol kg^–1, respectively. The detection limits (calculated as three times the standard error of the intercept/slope of the calibration line) were 0.27 μmol kg^–1 for NO₂^– + NO₃^– and 0.21 μmol kg^–1 for PO₄^3–.

Chl-a concentrations were estimated using a fluorometer (Trilogy, Turner Designs, San Jose, CA, USA) previously calibrated with pure Chl-a (Sigma, St. Louis, MO, USA) (Parsons et al., 1984).

Immediately prior to analysis, the samples were thawed, stained with SYBR Green I (Molecular Probes) for 15 min (Noble and Fuhrman, 1998), and then analyzed using an Accuri C6 flow cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) equipped with an air-cooled argon ion laser (488 nm, 15 mW). Bacteria were identified using the characteristics of their side light scatter and green fluorescence signals.

Riverine DOC concentration (μM C) was determined using the following equation, as described by Mathis et al. (2007):

where DOC_river is the initial DOC concentration in river runoff. In this study, a concentration of 350 μM C was used for DOC_river in the CS (i.e., stations 1–13) (Mathis et al., 2007). For the CBL and ESS, DOC_river was set to be 175 μM C, corresponding to zero-salinity (100% river water) DOC value, obtained from the relationship between DOC and sea ice meltwater-corrected S in these regions (Figure S1). Details underlying the choice of DOC_river in our study regions can be found in the publications by Jung et al. (2021b) and Jung et al. (2022).

The parameter N^* was calculated using the following equation reported by Gruber and Sarmiento (1997):

The underlying premise of N^* is that when marine organic matter is remineralized, the N/P atomic regeneration ratio is normally close to 16/1 (i.e., the Redfield ratio of N to P). The constant value 2.9 was used to make the global average of N^* for the ocean zero, and the value 0.87 was used to account for PO₄^3– released by the regeneration of organic matter during denitrification. The Atlantic waters entering the Arctic Ocean have positive N^* values owing to net nitrogen fixation in the North Atlantic, whereas that entering Pacific waters have negative values due to the significant denitrification that occurs within the shelf sediments in the Chukchi Sea (Devol et al., 1997; Codispoti et al., 2005; Codispoti et al., 2009; Nishino et al., 2013).

The θ–S diagram shows the presence of four major water masses in the western Arctic Ocean (Figure S2), consistent with previous studies (Codispoti et al., 2005; Nishino et al., 2013; Alkire et al., 2019): surface mixed layer (< 50 m depth), Pacific summer water (S = 31–32 psu, θ maximum, 50–100 m depth), Pacific winter water (S ≈ 33 psu, θ minimum, 100–150 m depth), and Atlantic water (S > 34 psu, maximum θ, below 150 m depth). In the CBL and ESS, the Atlantic-origin cold halocline water (S ≈ 34.2~34.5 psu, θ ≈ –1 °C) was also observed (Jung et al., 2021a).

Figure 2 shows the hydrographic conditions (S, θ, inorganic nutrients, N^*, and Chl-a) along a latitudinal transect from the CS to CBL (i.e., transect 1). The warmest waters (3.2 < θ < 10.9 °C) were observed in the upper 50 m of the southern CS (stations 1–4 and 10), which became shallow as they flowed into the northern CS and CBL. Compared to those in the CS (range: 29.9 < S < 33.0 psu, mean: 32.0 ± 0.7 psu), S values in the upper 50 m of the CBL (range: 27.2 < S < 31.8 psu, mean: 29.7 ± 1.8 psu) were lower, presumably due to river discharge and/or melting of sea ice. The upper halocline layer with salinities of 32–33.5 psu at depths between 50 and 200 m had high inorganic nutrient concentrations (mean: 1.4 ± 0.3 μmol kg^–1 for PO₄^3– and 9.4 ± 4.5 μmol kg^–1 for NO₂^– + NO₃^–) and low N^* (range: –10.4 to –7.6 μmol kg^–1, mean: –8.9 ± 0.9 μmol kg^–1), suggesting the characteristics of Pacific-origin water (Itoh et al., 2015; Danielson et al., 2017). The warm (θ > 0°C) and saline (S > 34 psu) Atlantic water lay beneath the upper halocline layer. The Chl-a concentrations ranged from 0.07 to 17.9 mg m^–3 (mean: 3.1 ± 4.0 mg m^–3), with the highest values in the CS.

The vertical distributions of S and θ along a longitudinal transect from the ESS to CBL (i.e., transect 2) showed a vertical stratification (Figure 3). The surface mixed layer (< 25 m) was characterized by relatively warm (range: −1.5 < θ < 5.2 °C, mean: 1.0 ± 2.2 °C) and fresh waters (range: 27.4 < S < 31.4 psu, mean: 29.4 ± 1.2 psu). Below the surface mixed layer, the Pacific summer water (S = 31–32 psu, θ maximum) was observed. The upper halocline layer (32 < S < 33.5 psu, at depths between 50 and 200 m) had higher concentrations of inorganic nutrients (mean: 1.6 ± 0.4 μmol kg^–1 for PO₄^3– and 12.4 ± 4.5 μmol kg^–1 for NO₂^– + NO₃^–) and lower N^* values (range: –12.9 to –4.5 μmol kg^–1, mean: –9.5 ± 1.8 μmol kg^–1) (Figures 3C−E). Interestingly, the upper halocline waters with salinities of 32–33.5 psu shoaled up to approximately 30 m at station 26. The Chl-a maximum was found at depth of 22 m in station 26 (Figure 3F). The densest water (S > 34 psu and θ ≈ –1 °C) observed at depths below 80 m in the west part of the ESS was associated with Atlantic-origin waters (Codispoti et al., 2009; Kondo et al., 2016), and its layer deepened toward the east.

The distributions of f_river along transects 1 and 2 are shown in Figures 4A, B. The f_river along transect 1 ranged from 0 to 0.18, with means of 0.06 ± 0.02 in the CS and 0.10 ± 0.05 in the CBL (Figure 4A). The f_river along transect 2 exhibited similar distribution as along transect 1, varying from 0 to 0.18 (Figure 4B). Overall, the f_river values in the upper 50 m of the CBL (mean: 0.13 ± 0.03) were higher than those in the ESS (mean: 0.09 ± 0.02). The highest f_river values were observed in the eastern parts of transect 2 (i.e., stations 15, 24–25), where the regions are in accord with the position of the edge of Beaufort Gyre associated with freshwater accumulation (Jung et al., 2021b).

The vertical distributions of riverine DOC in transects 1 and 2 were similar to those of f_river (Figures 4C, D). The riverine DOC concentrations along transect 1 ranged from 1.04 to 55.4 µM C (mean: 22.8 ± 11.0 µM C), with the highest values in the surface layer and decreasing with depth (Figure 4C). The mean concentration of riverine DOC in the upper 50 m of the CBL (27.1 ± 2.6 µM C) was higher compared to that in the CS (20.9 ± 6.0 µM C). The riverine DOC along transect 2 exhibited similar distribution as that along transect 1, varying from 0 to 31.2 µM C (mean: 13.2 ± 8.2 µM C) (Figure 4D).

Heterotrophic bacterial abundance along transects 1 and 2 ranged from 0.24 to 2.5 × 10⁶ cells mL^–1 and 0.11 to 1.8 × 10⁶ cells mL^–1, respectively (Figure 5). Along transect 1, the mean abundance of heterotrophic bacteria in the upper 50 m of the CBL (mean: 0.37 ± 0.10 × 10⁶ cells mL^–1) was lower than that in the CS (mean: 1.3 ± 0.54 × 10⁶ cells mL^–1). Heterotrophic bacterial abundance in the surface layer of the CBL along transect 2 (mean: 0.36 ± 0.18 × 10⁶ cells mL^–1) was lower than that in the ESS (mean: 0.65 ± 0.42 × 10⁶ cells mL^–1).

Three different fluorescent components were identified by the PARAFAC analysis, based on the results from core consistency and split-half validation, and are henceforth referred to as components 1 to 3 (Figure 6). The three components were compared to those identified previously from the OpenFluor database with Tucker’s congruence coefficients exceeding 0.95 (Murphy et al., 2014) and by visual inspection (Table S1) (Stedmon et al., 2007; Stedmon et al., 2011; Dainard and Guéguen, 2013; Catalá et al., 2015; Gonçalves-Araujo et al., 2015; Gonçalves-Araujo et al., 2016; Gao and Guéguen, 2017; Brogi et al., 2019; Kim et al., 2020). Component 1 (C1) showed maximum Ex wavelengths below 250 nm and at 310 nm, and a maximum Em wavelength at 400 nm (Figures 6A, D), whereas component 2 (C2) showed Ex maxima at 260 nm and 370 nm, and an Em maximum at 473 nm (Figures 6B, E). The C1 and C2 were categorized as the humic-like FDOM, since they had broader Ex spectra and Em maxima above 380 nm. The results obtained from the OpenFluor database for C1 were somewhat contradictory; some indicated C1 to be similar to the terrestrial humic-like component (Stedmon et al., 2007; Gonçalves-Araujo et al., 2015) while others showed C1 to have the characteristics of marine humic-like component in the visible region (Catalá et al., 2015; Gao and Guéguen, 2017; Kim et al., 2020). Therefore, C1 was categorized as a mixture of the terrestrial and marine humic-like components, described earlier as peaks A and M by Coble (1996) through visual inspection (Ishii and Boyer, 2012). C2 has fluorescence properties similar to that of terrestrial humic-like component (Stedmon et al., 2011; Dainard and Guéguen, 2013; Gonçalves-Araujo et al., 2016; Kim et al., 2020). It was characterized as a mixture of the terrestrial humic-like components in the UV and visible regions, described earlier as peaks A and C (Coble, 1996). Component 3 (C3) displayed an Ex maximum at 280 nm and an Em maximum at 341 nm (Figures 6C, F). It had similar spectral characteristics as the tryptophan-like component (Stedmon et al., 2011; Brogi et al., 2019).

The surface distributions of the humic-like and tryptophan-like FDOM and S_275–295 are shown in Figure 7. The fluorescence intensities of a mixture of terrestrial and marine humic-like C1 and terrestrial humic-like C2 varied from 0.018 to 0.063 R.U. and 0.009 to 0.028 R.U., respectively (Figures 7A, B). Notably, the CBL exhibited lower intensities of humic-like FDOM (mean: 0.027 ± 0.006 R.U. for C1 and 0.011 ± 0.003 R.U. for C2) compared to the CS (mean: 0.036 ± 0.018 R.U. for C1 and 0.015 ± 0.006 R.U. for C2) and ESS (mean: 0.036 ± 0.004 R.U. for C1 and 0.016 ± 0.001 R.U. for C2). The tryptophan-like C3 exhibited values ranging from 0.011 to 0.026 R.U., with the highest concentrations observed in the CS (Figure 7C). The S_275–295 values varied between 0.006 and 0.041 nm^–1, with the highest values observed in the eastern stations of the CBL (i.e., stations 15–18), ranging from 0.031 to 0.041 nm^–1 (Figure 7D). The lowest S_275–295 value (0.006 nm^–1) was observed in conjunction with high fluorescent intensity values of the humic-like FDOM at station 26. Notably, this station coincided with the location where shoaling of the upper halocline layer was observed (Figures 3A–E). In addition, elevated S_275–295 values and reduced fluorescence intensities of humic-like FDOM were observed in the surface water of the eastern stations in the CBL, in contrast to the CS and ESS.

The vertical distributions of C1 and C2 fluorescence intensities along transects 1 and 2 are shown in Figures 8A–D. The intensities of C1 and C2 along transect 1 ranged from 0.018 to 0.063 R.U. and 0.010 to 0.028 R.U., respectively (Figures 8A, C). Intensities of the humic-like FDOM were low in the surface layer and high in the upper halocline layer. Along transect 2, C1 and C2 values varied from 0.022 to 0.066 R.U. and 0.010 to 0.031 R.U., respectively, with the highest intensities in the upper halocline layer and the lowest intensities at the surface. Interestingly, higher intensities of humic-like FDOM (C1 and C2) than that in surrounding waters were found in the upper 30 m at station 26 along transect 2, where Chl-a was observed to be maximum (Figure 3F).

The tryptophan-like C3 showed different patterns compared to the humic-like FDOM (C1 and C2) (Figures 8E, F). The intensities of tryptophan-like FDOM along transects 1 and 2 ranged from 0.003 to 0.047 R.U. and 0.003 to 0.041 R.U., respectively, showing a pattern of decreasing intensities with depth. Fluorescence intensities of the tryptophan-like FDOM in the upper 50 m (mean: 0.017 ± 0.007 R.U. in transects 1 and 2) were higher than that at depths below 50 m (mean: 0.010 ± 0.005 R.U. in transect 1 and 0.007 ± 0.001 R.U. in transect 2). Furthermore, distributions of the tryptophan-like FDOM showed an overall south-north gradient for transect 1 and an east-west gradient for transect 2. Compared to those in the CS and ESS, the intensities of tryptophan-like FDOM were higher at depths of approximately 50 m in the CBL, where the f_river and riverine DOC levels were maximum (Figure 4). Contrary to previously published results, which suggested that tryptophan-like FDOM is produced from biological activities (Determann et al., 1998; Moran et al., 2000; Stedmon et al., 2003; Romera-Castillo et al., 2011; Chen et al., 2017; Chen et al., 2018), the vertical distributions of the tryptophan-like FDOM along transects 1 and 2 were different from those of Chl-a (Figures 2F, 3F) and heterotrophic bacterial abundance (Figure 5).

To explore the origin of tryptophan-like FDOM (C3), we examined the correlation of tryptophan-like FDOM with biological factors, such as Chl-a and heterotrophic bacterial abundance, as well as river water parameters, including f_river and riverine DOC (Figure 9). There was no significant correlation between tryptophan-like FDOM and biological factors (Figures 9A, B), which implied that the tryptophan-like FDOM may have originated from processes other than biological activities. Additionally, no significant relationship was found between the tryptophan-like FDOM and f_{sea ice melt} (Figure S3A), suggesting that the contribution of sea ice meltwater to the tryptophan-like FDOM is negligible. On the other hand, the tryptophan-like FDOM showed significant positive relationships with f_river and riverine DOC (Figures 9C, D), indicating a strong association between tryptophan-like FDOM and riverine inputs. Moreover, a negative relationship was found between the tryptophan-like FDOM and f_seawater (Figure S3B) in the study regions, further supporting the association between tryptophan-like FDOM and riverine contributions. Furthermore, the magnitude of correlation coefficients between the tryptophan-like FDOM and river water parameters (r = 0.55 for the CS, r = 0.51 for the CBL, and r = 0.85 for the ESS) was different across regions (Figures 9C, D). Plots between the tryptophan-like FDOM intensities and river water parameters in the ESS were scattered over a narrower range than those in the CS and CBL, indicating that the influence of river water on tryptophan-like FDOM is much stronger in the ESS than in the CS and CBL.

The upper 50 m of the water column in the Arctic Ocean is generally depleted of nutrients (PO₄^3– < 1 μmol kg⁻¹ and NO₂^– + NO₃^– < 3 μmol kg⁻¹) during summer (Lee et al., 2012; Hioki et al., 2014; Coupel et al., 2015). However, higher concentrations of inorganic nutrients (up to 2 μmol kg⁻¹ for PO₄^3– and 15 μmol kg⁻¹ for NO₂^– + NO₃^–) were observed in the upper 50 m of the ESS along transect 2 (Figures 3C, D), resulting in phytoplankton bloom (Figure 3F). The distributions of S and θ along transect 2 showed shoaling of the upper halocline layer, with doming of the isopycnals (Figures 3A, B). Moreover, the Atlantic-origin cold halocline water shoaled up to a depth of 80 m (Figure 3), which was shallower than the depths of 140–150 m where the Atlantic water is usually found (Jung et al., 2021a). This implied that shoaling of both the upper halocline layer and Pacific-origin water by the intrusion of Atlantic-origin cold halocline water triggered the supply of inorganic nutrients into the surface layer, thereby resulting in phytoplankton bloom. This was consistent with the result reported by Jung et al. (2021a), which suggested that the Atlantic-origin cold halocline water extended as far as the region of Chukchi Sea due to the strong cyclonic circulation, resulting in shoaling of the upper halocline layer.

Interestingly, the humic-like FDOM (C1 and C2) was remarkably high in the upper halocline layer with low N^* values (Figures 2, 3, and 8). Since N^* is known as the deviation of nitrate from the stoichiometric ratio of nitrate to phosphate (Gruber and Sarmiento, 1997), it is widely used to estimate the excess nitrate (e.g., nitrogen fixation) or nitrate deficit (e.g., denitrification) relative to phosphate. A positive N^* value corresponds to high nitrogen input while a negative N^* value is indicative of nitrogen deficiency from sediment denitrification and/or water column denitrification. Extremely low N^* values (< –10 µmol kg^-1) are rare, occurring in only approximately 3% of global measurements (Deutsch and Weber, 2012). The Bering and Chukchi seas are known as hotspots of sedimentary denitrification, resulting in an extremely low N^* of approximately –10 µmol kg^-1 (Devol et al., 1997; Tanaka et al., 2004; Yamamoto-Kawai et al., 2006; Chang and Devol, 2009; Brown et al., 2015). Therefore, the layer of high humic-like FDOM intensities with the lowest N^* suggested that the humic-like FDOM was supplied from the shelf sediments. This finding was consistent with that of previous studies, which showed that the humic-like FDOM is released from shelf sediment during early diagenesis and then spread laterally along the upper halocline layer of the Arctic Ocean (Hioki et al., 2014; Chen et al., 2018).

In addition, the high intensities of humic-like FDOM in the surface layer (< 30 m) at station 26 (Figures 8B, D) coincided with the region where shoaling of the upper halocline layer was substantial. As reported in previous studies (Polyakov et al., 2017; Ardyna and Arrigo, 2020; Polyakov et al., 2020; Jung et al., 2021a), the impacts of climate change on the atmospheric circulation have significant implications on ocean circulation in the western Arctic Ocean. Observation and modeling results suggested the shift of ocean circulation to a more cyclonic state in the 1990s (Steele and Boyd, 1998; Johnson and Polyakov, 2001). In addition, Jung et al. (2021a) reported that the enhanced cyclonic winds in 2017 led to the intrusion of Atlantic-origin cold halocline water from the Eurasian seas into the Chukchi Sea along the shelf slope, resulting in shoaling of the Pacific-origin water into the euphotic zone. As a result of this intrusion of Atlantic-origin cold halocline water into the Chukchi Sea, high intensities of the humic-like FDOM were supplied to the euphotic zone with shoaling of the upper halocline layer, thereby causing an exposure of the humic-like FDOM to sunlight.

Generally, tryptophan-like FDOM is known to be produced by biological activities, such as direct exudation from phytoplankton (Stedmon and Markager, 2005b; Romera-Castillo et al., 2010; Romera-Castillo et al., 2011), release during zooplankton grazing (Urban-Rich et al., 2004; Urban-Rich et al., 2006), production via heterotrophic bacterial processes (Moran et al., 2000; Rochelle-Newall and Fisher, 2002; Romera-Castillo et al., 2011; Chen et al., 2018), and photodegradation of humic-like DOM (Romera-Castillo et al., 2013). However, the tryptophan-like FDOM (C3) showed no significant relationships with Chl-a and heterotrophic bacterial abundance in this study (Figures 9A, B), suggesting that it could be associated with other sources and/or processes rather than with biological activities. Interestingly, higher fluorescence intensities of tryptophan-like FDOM were found in the CBL along transects 1 and 2 (Figures 8E, F), where the f_river and riverine DOC were maximum (Figure 4) and Chl-a and heterotrophic bacterial abundance were low (Figures 2F, 3F, and 5). Furthermore, intensity of tryptophan-like FDOM was significantly positively correlated with f_river and riverine DOC (Figures 9C, D), indicating that riverine input could be a source of tryptophan-like FDOM. This was consistent with the previous results observed in the Canada Basin (Dainard et al., 2019) and the East Siberian Sea (Drozdova et al., 2022). River runoff supplies large amounts of terrestrial DOM to the Arctic Ocean (Gordeev et al., 1996; Amon et al., 2012; Holmes et al., 2012; Anderson and Amon, 2015). For example, Opsahl et al. (1999) estimated that approximately 5–33% of DOM on the Arctic Ocean surface is of terrigenous origin. Terrestrial DOM pools consist of complex mixtures of compounds of varying ages (Raymond and Spencer, 2015). In addition, the PARAFAC-validation components cannot be attributed to only pure compounds, but also to a group of compounds that exhibit similar fluorescence properties as of pure compounds (Gao and Guéguen, 2017). This suggested that the FDOM components are a complex mixture of compounds that have overlapping absorption and emission spectra, with no single compound dominating. Stedmon and Markager (2005b) reported that tryptophan-like FDOM is linked to the degradation of terrestrial DOM. Previous studies (Maie et al., 2007; Hernes et al., 2009; Yamashita et al., 2017) had revealed that both propylphenol monomers and tannins, typical biomarkers of vascular plant, fluoresce in the region of the EEMs that is traditionally defined as the tryptophan-like region. In addition, Maie et al. (2007) had separated the tryptophan-like FDOM into two compounds using size exclusion chromatography, namely proteinaceous materials and phenolic moieties of humic substances. The finding showed that tryptophan-like FDOM consists of a mixture of compounds with different chemical structures, namely labile proteinaceous substances and phenolic moieties in the humic-like substances. Phenolic compounds are mainly derived from the breakdown of lignin, which is regarded as a biomarker for terrestrial organic matter (Saiz-Jimenez and Leeuw, 1986). Thus, higher intensities of tryptophan-like FDOM in the CBL suggested that tryptophan-like FDOM could be mainly derived from river discharge.

Further, the results of this study showed that the influences of f_river and riverine DOC on tryptophan-like FDOM are region-dependent (Figures 9C, D). The high correlation coefficients between tryptophan-like FDOM and f_river and riverine DOC in the ESS indicated that tryptophan-like FDOM in the ESS could be mainly regulated by a riverine source. In contrast, the low correlation coefficients in the CS and CBL suggested that there is an additional mechanism for regulating the relationship between tryptophan-like FDOM and f_river and riverine DOC in the CS and CBL. A possible explanation for this difference could be in terms of biological activities. Along transect 2 across the ESS and CBL, heterotrophic bacterial abundance was higher in the ESS than in the CBL. Tryptophan-like FDOM is significantly correlated with the percent of biodegradable organic carbon (Fellman et al., 2009; Hood et al., 2009) that is rapidly consumed by bacteria in the surface waters (Rich et al., 1996; Ogawa et al., 2001; Mathew et al., 2021). Furthermore, Lønborg et al. (2015) reported that the tryptophan-like FDOM decayed by 30% after 72 h of incubation in the darkness. Therefore, the labile proteinaceous substance may be inferred to be consumed by the high abundance of heterotrophic bacteria in the ESS, while the phenolic moieties originating from river water remained, resulting in the high correlation coefficients in the relationships between tryptophan-like FDOM and river water parameters (i.e., f_river and riverine DOC) in the ESS.

The DOM cycling in the Arctic marine system can be impacted by various environmental changes, such as increase in river discharge (Peterson et al., 2002; Holmes et al., 2021), decline in sea ice coverage (Cavalieri and Parkinson, 2012; Meier et al., 2021), modifications of water mass structures resulting from changes in the ocean circulation system (Polyakov et al., 2017; Polyakov et al., 2020), and increase in net primary production (Ardyna and Arrigo, 2020). For example, the reduction in sea ice coverage has led to an increase in open water area and prolonged duration of sea ice-free conditions, allowing for increased penetration of solar radiation into the Arctic Ocean (Nicolaus et al., 2012). In addition to the reduced sea ice coverage, the increase in river discharge can lead to a freshening of the Arctic Ocean, resulting in a longer residence time caused by stratification, which provides favorable conditions for the photochemical processes of DOM (Osburn et al., 2014). Previous studies have revealed that humic-like FDOM is susceptible to photodegradation (Vecchio and Blough, 2002; Jørgensen et al., 2014; Logvinova et al., 2015). Photodegradation of high-molecular-weight DOM typically leads to a decrease in the average molecular weight of DOM, consequently resulting in an increase in S_275–295 (Helms et al., 2008; Lu et al., 2016). Helms et al. (2008) reported an increase in S_275–295 values during the course of irradiation, indicating a decrease in the molecular weight of DOM due to the photodegradation of high-molecular-weight DOM. Similarly, Lu et al. (2016) observed a gradual increase in S_275–295 values from 0.014 nm^–1 on day 0 to 0.027 nm^–1 on day 62 as the irradiation progressed. Thus, S_275–295 values have been widely used as a reliable proxy of DOM molecular weight and photochemical processes (Helms et al., 2008; Fichot and Benner, 2012; Granskog, 2012; Granskog et al., 2012; Dainard and Guéguen, 2013; Helms et al., 2013; Yamashita et al., 2013b; Lee et al., 2016; Shen et al., 2016a; Johnston et al., 2018; Hirawake et al., 2021). In this study, we found that the humic-like FDOM derived from the shelf sediments was supplied to the euphotic zone through shoaling of the upper halocline layer due to the intrusion of Atlantic-origin cold halocline water into the CBL, thus exposing the humic-like FDOM to sunlight (Figures 8B, D). This claim can be supported by the observation of the lowest S_275–295 value, along with high fluorescent intensity values of the humic-like FDOM in the surface water at station 26 (Figures 1, 7), where shoaling of the upper halocline layer was observed (Figures 3A–E). The occurrence of the lowest S_275–295 value, coupled with the high fluorescent intensity values of the humic-like FDOM in the surface water of the CBL, provides strong evidence for the transport of high-molecular-weight DOM to the surface water as a consequence of shoaling of the upper halocline layer. Moreover, we observed elevated S_275–295 values and reduced fluorescence intensities of the humic-like FDOM in the surface water of the eastern stations in the CBL, in contrast to the CS and ESS (Figures 7A, B, D). Given the susceptibility of high-molecular-weight DOM to photodegradation (Vecchio and Blough, 2002; Jørgensen et al., 2014; Logvinova et al., 2015), these results strongly support the occurrence of DOM photodegradation, which is likely enhanced in the highly stratified surface waters with a prolonged residence time in the region influenced by the Beaufort Gyre. Similarly, Shen et al. (2016a) reported elevated S_275–295 values in the surface waters of the Beaufort Gyre, suggesting significant DOM removal through photodegradation, which was facilitated by a longer residence time in the region. In addition, the conversion of humic-like FDOM to low-molecular-weight DOM through photodegradation increases its bioavailability (Tranvik and Bertilsson, 2001; Logvinova et al., 2015; Wilske et al., 2020). This process can also contribute to enhancing bacterioplankton productivity (Logvinova et al., 2015). Previous studies (Nakayama et al., 2011; Hioki et al., 2014) had revealed that inorganic nutrients and dissolved iron, as well as the humic-like FDOM, are supplied from shelf sediment and transported through the upper halocline layer in the Arctic Ocean. Considering that the supply of sufficient dissolved iron and inorganic nutrients to the euphotic zone can lead to an increase in primary production, the transport of these constituents to the euphotic zone by shoaling of the upper halocline layer, coupled with the transformation of the humic-like FDOM into low-molecular-weight DOM through photodegradation, could result in the consumption of humic-like FDOM by microbial and/or photochemical degradation along with fresh DOM produced by primary production (Shen and Benner, 2018). Furthermore, photodegradation of DOM can lead to the production of inorganic carbon forms, CO and CO₂ (Mopper et al., 1991; Helms et al., 2008). Consequently, our results highlight that the transport of the humic-like FDOM to the euphotic zone due to a change in ocean circulation pattern has the potential to impact the marine carbon and biogeochemical cycles in the western Arctic Ocean.

As the Arctic Ocean continues to warm, the projected increase in river discharge (Peterson et al., 2002; Holmes et al., 2021) will introduce phenolic compounds directly leached from terrestrial plant materials (Benner et al., 1990; Hernes et al., 2001) into the Arctic Ocean. The mobilization of these plant-derived materials has increased by up to 6% from 1985 to 2004 (Feng et al., 2013) and is expected to increase further under the current warming trends (Amon et al., 2012). In addition, Bertosio et al. (2022) reported that the Beaufort Gyre, the major reservoir of Arctic freshwater, extended northward, and freshwater content increased in the central Arctic from 2012 onward due to changes in atmospheric circulation patterns. Thus, additional tryptophan-like FDOM (i.e., phenolic moieties) (Figure 9) could be introduced into the Arctic surface waters via river discharge and carried further north. Moreover, Müller et al. (2018), who reported the effects of terrestrial-derived DOM on marine microbial community composition, suggested that the addition of terrestrial-derived DOM to seawater samples induces changes in bacterial community composition, leading to an increase in bacterivorous grazing by small protists. This implies that the supply of tryptophan-like FDOM via river discharge may have an impact not only on bacteria but also on higher trophic levels of the microbial food webs indirectly.

Furthermore, a long-term study utilizing an ocean color algorithm parameterized for the Arctic Ocean revealed that primary production, particularly along the interior shelf break, increased by 57% between 1998 and 2018 as a result of melting sea ice and the consequent amplification of phytoplankton biomass (Lewis et al., 2020). The increase in Arctic primary production has the potential to further enhance the supply of tryptophan-like FDOM (i.e., labile proteinaceous substances). This was supported by DeFrancesco and Guéguen (2021), who reported a significant increase in protein-like FDOM in the upper polar mixed layer of the central Canada Basin between 2007 and 2017. Because labile tryptophan-like FDOM is strongly correlated with the percent of biodegradable organic carbon (Fellman et al., 2009; Hood et al., 2009) that is rapidly consumed by bacteria (Rich et al., 1996; Ogawa et al., 2001; Hansell, 2013; Mathew et al., 2021), the increased supply of labile tryptophan-like FDOM is expected to provide a significant amount of biodegradable organic carbon. This, in turn, can induce changes in the quantity and quality of DOM in the Arctic Ocean.

FDOM, a sub-fraction of chromophoric DOM (CDOM), absorbs visible radiation and consequently contributes to the heating of surface waters (Kirk, 1988; Hill, 2008). Hill (2008) reported that elevated CDOM concentrations can lead to the trapping of radiant heat in the mixed layer of the water column. Thus, the enhanced supply of humic-like FDOM originating from the shelf sediment through shoaling of the upper halocline layer, as well as tryptophan-like FDOM derived from marine primary production and river discharge (i.e., both labile proteinaceous substances and phenolic moieties) to the euphotic zone can also affect the heat budget in the Arctic Ocean. However, it is worth noting that these FDOM components are subject to processes such as photodegradation and bacterial consumption, which can alter their concentrations and influence their impact on the heat budget. Nevertheless, the potentially photobleaching-resistant fractions of humic-like FDOM, along with the less photolabile tryptophan-like FDOM, can enhance the absorption of solar energy and trap additional heat in the mixed layer, thus potentially affecting the heating budget of Arctic surface waters (Yang et al., 2020). Thus, our results suggest that the increased supply of humic-like and tryptophan-like FDOM to the euphotic zone could potentially impact the dynamics of marine food webs, DOM cycling, and the heating budget in the western Arctic Ocean.