The conservative behavior of dissolved organic carbon in surface waters of the southern Chukchi Sea, Arctic Ocean, during early summer

The spatial distribution of dissolved organic carbon (DOC) concentrations and the optical properties of dissolved organic matter (DOM) determined by ultraviolet-visible absorbance and fluorescence spectroscopy were measured in surface waters of the southern Chukchi Sea, western Arctic Ocean, during the early summer of 2013. Neither the DOC concentration nor the optical parameters of the DOM correlated with salinity. Principal component analysis using the DOM optical parameters clearly separated the DOM sources. A significant linear relationship was evident between the DOC and the principal component score for specific water masses, indicating that a high DOC level was related to a terrigenous source, whereas a low DOC level was related to a marine source. Relationships between the DOC and the principal component scores of the surface waters of the southern Chukchi Sea implied that the major factor controlling the distribution of DOC concentrations was the mixing of plural water masses rather than local production and degradation.

Scientific RepoRts | 6:34123 | DOI: 10.1038/srep34123 autochthonous DOC and dissolved organic nitrogen (DON) to the surface water of the western Arctic Ocean have been evaluated based on the spatial distributions combined with salinity and δ 18 O 13,14 . The contribution of autochthonous DOM in the upper Arctic Ocean has also been described based on the temporal or spatial distributions of biochemical compounds [15][16][17] .
Previous studies have successively clarified that several sources, autochthonous production, and removal are possibly important for controlling the DOM distribution in the Arctic Ocean, as mentioned above. These findings were often based on temporal changes in the DOM and/or the spatial distributions of DOM and salinity (and δ 18 O) with a two or three end-member analysis. However, temporal or two/three end-member analyses may not be sufficient to evaluate the major factor controlling the DOM distribution along the margins of the Arctic Ocean where plural water masses, i.e., at least two freshwater sources and several seawater sources, possibly contribute 4,7 . Therefore, a novel approach that is independent of physical parameters for water mass identification (i.e., temperature, salinity, and δ 18 O) is necessary to evaluate the major factor controlling the DOM distribution in the regions where a number of seawater sources are evident.
In this study, we determined the spatial distribution of the DOC concentrations and DOM optical properties in surface waters of the southern Chukchi Sea during the early summer of 2013. Several water masses can be discerned in the southern Chukchi Sea, but the DOC concentrations in specific water masses have not been well documented previously 4,18 . The water flow through the Bering Strait is northward during most of the year 19 and is an important source of heat, freshwater, and nutrients into the Arctic Ocean [20][21][22][23][24] . The Alaskan Coastal Water (ACW), the Bering Shelf Water (BSW), and the Anadyr Water (AW) enter from the Bering Sea through the Bering Strait, with the ACW on the east, the BSW in the middle, and the AW on the west 23,24 . The Siberian Coastal Current occasionally flows southward through the Bering Strait, primarily in the fall and winter 25 . Based on the relationship between the DOC concentration and DOM quality determined by optical analyses, the conservative behavior of the DOC was found to be the major factor controlling the DOC distribution in the southern Chukchi Sea during the early summer of 2013.

Results and Discussion
Water masses in the southern Chukchi Sea. In the early summer of 2013, the warmer, less saline, lower nutrient concentration ACW 23,24 , also known as the Eastern Chukchi Summer Water (ECSW) 20 , was distributed to the north of Cape Lisburne (Table 1; Supplementary Fig. S1; Fig. 1). The ACW is known to flow along the Alaskan coast and is affected by Alaskan rivers 23,24 . The AW and the BSW, which can be characterized as saline and having higher nutrient concentration compared to the ACW 23,24 , were distributed in the vicinity of the Bering Strait (Table 1; Supplementary Fig. S1; Fig. 1). The AW was distinguished from the BSW by temperature and salinity (Table 1; Supplementary Fig. S1; Fig. 1) because the AW can be characterized as cooler and more saline compared to the BSW 23,24 . It has been noted that a gradual interface between the AW and the BSW promotes the formation of a combined water mass 23 , and the AW and the BSW have been defined as one water mass, called the Western Chukchi Summer Water (WCSW) 20 ; thus, the AW and the BSW identified in this study possibly affected each other.
Sea-ice melting water is another source of freshwater for the upper water column of the southern Chukchi Sea, particularly at the sea-ice edge regions 23 . The other water mass distributed in the Chukchi Sea during summer is the Pacific Winter Water (PWW) 26,27 . The PWW is formed during sea-ice formation and thus is characterized as extremely cold, dense, and high in nutrients due to the supply from sediments. Large negative N* in the PWW indicates the influence of sedimentary denitrification 28 . In the present study, the PWW was distributed to the northeast of Cape Icy (Table 1; Supplementary Fig. S1; Fig. 1). The temperature and salinity of the PWW defined in this study were higher and lower than those of the previously defined PWW, respectively 26,27 , indicating that this water mass was a "modified" PWW, with mixing of the upwelled PWW and the surface water in the region. Other water masses (others), which were not identified from physicochemical parameters (i.e., temperature, salinity, and nutrient conditions), distributed across the southern Chukchi Sea during the early summer, 2013 ( Fig. 1), could be considered as a mixture of specific water masses (i.e., ACW, AW, BSW, PWW, and sea-ice melt water).
Distributions of DOC concentrations and DOM optical parameters in surface waters of the southern Chukchi Sea during early summer, 2013. The DOC concentration ranged from 62 to 98 μ MC in the study area and was similar to, or slightly lower than, values previously observed in the Bering Strait 12,18 and the Bering Shelf 29 . The distributional pattern of the DOC was not uniform, but it had a spatial variability related to the distribution of specific water masses (Fig. 2a) Table 1. Characteristics of specific water masses. n.d. = not determined. *Salinity was measured for different pieces of sea ice blocks from DOM analyses (n = 10). and the lowest DOC concentration was found in the AW and the BSW, near the Bering Strait, and in the central Chukchi Sea, near the ice edge ( Fig. 2; Table 1). The lowest DOC concentration was found in sea ice, even though the range was relatively large between the two samples (Table 1). Similar to the DOC-salinity relationship found for the saline waters of the Chukchi Sea 18 , the DOC concentration did not linearly correlate with salinity ( Fig. 2b) or temperature (R 2 = 0.15, p < 0.001, n = 83). Distribution patterns of optical properties of the DOM (DOM quality) were also not uniform. The spectral slope coefficient between 275 nm and 295 nm (S 275-295 ; Table 1; Supplementary Fig. S2) 30 , a tracer of terrigenous DOC 1 , had a similar range, as previously observed in the Chukchi Sea 1 and the Bering Shelf 29 . The range of S 275-295 observed in this study indicated a less variable and a small contribution of terrigenous DOC in the southern Chukchi Sea compared to the Eurasian margins of the Arctic Ocean 1 . The values of specific UV absorbance (SUVA 254 ; Table 1; Supplementary Fig. S2), an index for the aromaticity of DOM 31 , were lower than those found in the Yukon River basin 32 . Two and one fluorescent components, identified by excitation-emission matrix fluorescence (EEM) combined with parallel factor analysis (PARAFAC), could be categorized as humic-like components (C1 and C2) and a protein-like component (C3), respectively ( Supplementary Fig. S3), based on a comparison with previous studies of the Arctic Ocean [33][34][35] and the eastern Bering Sea 29 . The distributions of the relative abundance of humic-like C1 (%C1) and protein-like C3 (%C3) were similar to those of the SUVA 254 and S 275-295 , respectively ( Supplementary Fig. S2). Similar to the DOC concentrations, the optical parameters were different among the specific water masses ( Table 1). The ACW was characterized as having the lowest S 275-295 and %C3 and the highest SUVA 254 and %C1, whereas the highest S 275-295 and %C3 and the lowest SUVA 254 and %C1 were evident in the AW. The optical characteristics of the DOM in the PWW and BSW were in the middle ranges and close to those of the ACW and AW, respectively. The SUVA 254 values in both sea-ice samples were extremely low compared to those in surface waters, and the %C1 and %C3 of the sea ice were lower and higher compared to the surface waters, respectively (Table 1), indicating low aromaticity but richness of the protein-like component in the sea-ice DOM.
Even though the SUVA 254 was weakly correlated with salinity, other optical parameters were not correlated with salinity (Fig. 3), implying that salinity cannot be used to evaluate the conservative nature of the DOM quantity and quality in the Chukchi Sea, as previously observed for the Arctic margins 7,9,13 . The optical parameters were also not correlated with temperature (R 2 < 0.02, p > 0.05, n = 83 for S 275-295 , %C1, %C2, and %C3; R 2 = 0.08, p = 0.009, n = 83 for SUVA 254 ). In contrast, significant correlations were evident between the DOC concentrations and DOM optical parameters, except for %C2 (Fig. 4). Such correlations imply that the major factors controlling the DOC concentration and optical properties of the DOM (DOM quality) were similar in the surface waters of the southern Chukchi Sea during the early summer.
Dynamics of DOC in the surface waters of the southern Chukchi Sea during the early summer, 2013. Principal component analysis (PCA) was conducted using optical parameters (i.e., S 275-295 , SUVA 254 , %C1, %C2, and %C3) of all surface water and sea ice samples to assess the comprehensive DOM quality in terms of optical properties. The first and second principal components explained 71% and 15% of the variability, respectively. Figure 5a shows the property-property plot between the first and second factor loadings. SUVA 254 , %C1, and %C2 concurrently showed positive first factor loading. S 275-295 and %C3 showed negative first factor loading. The terrigenous DOM was categorized by low S 275-295 1 , high SUVA 254 3 , and the dominance of humic-like fluorophores 34,36 , whereas the marine DOM was characterized by the opposite trends. Thus, the first principal component represented the DOM sources; namely, a positive value indicated a terrigenous origin, whereas a negative value implied a marine origin. Figure 5b shows the relationship between the DOC concentration and the first principal component score for specific water masses. The relationship had differences in the DOM quantity and quality among the Pacific originated waters. The ACW was characterized by a high DOC concentration with terrigenous characteristics, whereas the AW and BSW had low DOC concentrations with marine features. This result implied that the terrigenous DOM from Alaskan rivers, e.g., the Yukon and Kuskokwim Rivers, contributed to the ACW, as indicated in previous studies 26,29 , whereas the contribution of terrigenous DOM from Siberian rivers, e.g., the Anadyr River, to the AW was possibly less important. The PWW was characterized by a higher DOC concentration and a greater contribution of terrigenous DOM compared to the AW and the BSW, implying that the DOM with a terrigenous feature may be derived from sediments during sea-ice formation 34,37 . The DOM in sea ice was characterized by the lowest DOC concentrations, with marine characteristics. Interestingly, there was a significant linear relationship between the DOC concentration and the first principal component score of the specific water masses, including sea ice (R 2 = 0.85, p < 0.001, n = 20; Fig. 5b). The relationship showed that high levels of DOC were related to terrigenous characteristics, whereas low levels of DOC were related to marine characteristics. Furthermore, this observation implies that the major factor controlling the distributions of the DOC concentration and DOM quality in other water masses that could not be identified from the physicochemical parameters (Fig. 1) can be evaluated from the linear relationship. Figure 5c shows the scatter plot of the DOC concentration and first principal component score of the other water masses. The regression line and the 95% prediction intervals of the regression obtained from specific water masses (Fig. 5b) are also shown in Fig. 5c. Almost all the data points of the other water masses were distributed between the 95% prediction intervals, indicating that a major factor controlling the distribution of the DOC concentration and DOM quality was the mixing of specific water masses, namely, the ACW, AW, BSW, PWW, and sea-ice melt water, in the surface waters of the southern Chukchi Sea during the early summer of 2013.
Previous studies have suggested that the photochemical process bleaches and alters the optical properties of DOM in the upper waters of the Arctic Ocean 7,11,34,38,39 , whereas others have noted that photobleaching of DOM is minor due to the specific conditions of the Arctic Ocean, namely, ice cover, low sun angle, and the strong attenuation of UV radiation by particles and CDOM 3,40-42 . The conservative behavior of DOC and DOM quality ( Fig. 5b,c) implied that photobleaching of DOM was minor for the surface waters of the southern Chukchi Sea during the early summer. However, it should be noted that the photobleaching of DOM possibly exhibits seasonality with the retreat of sea-ice 29 . The contribution of autochthonous biomolecules (amino acids, amino sugars, and carbohydrates) to the DOM in surface waters of the western Arctic Ocean during the late summer (late July-August) has been previously described 15,17,43 . In the present study, EEM-PARAFAC identified a protein-like component ( Supplementary  Fig. S3); however, the linear relationship between the DOC and the first principal component score was clear (Fig. 5b,c). This result suggested that local production and/or consumption of the protein-like component did not substantially affect the DOC distribution, even though active cycling (i.e., production and consumption) of labile DOM possibly occurred locally. Because the relative abundance of the protein-like component in the AW and BSW was greater than those in the ACW and the PWW (Table 1), the semi-labile protein-like component produced in the Bering Sea was possibly conservatively distributed in the southern Chukchi Sea during the early summer of 2013. The contribution of microbial CDOM has also been described in the western Arctic Ocean 34,39,44 . It has been well documented that the microbial processing of DOM generates new compounds into the environment 45,46 , and part of the CDOM could be showing humic-like fluorescent characteristics and be refractory, such that it can stay in the water for a long time and contribute to carbon sequestration in the ocean 47,48 . Such microbial processes should occur in the Arctic Ocean but might be slow due to the low temperature. Even if the production rate of the refractory DOM is low, it accumulates over time, thus contributing to the DOC concentration in the Arctic Ocean as a marine end-member.
The novel approach used in this study clarified that conservative mixing, rather than local production and degradation, was an important factor controlling the DOM distribution in the surface waters of the Chukchi Sea during the early summer, 2013. Levels of humic-like fluorophore can be determined by in situ sensors 34,49,50 . The conservative behavior of a humic-like component implies that the humic-like fluorescence intensity, determined by in situ sensors, could be used as one of the physicochemical parameters to determine the water mass, which in turn affects the biological production in the Chukchi Sea 50 . The degradability of terrigenous DOM in the Arctic Ocean is controversial, and the non-conservative distribution of DOC has been noted in the Arctic Ocean, even along the river-influenced margin 6,8,9 . Such a discrepancy in the dynamics of DOC can be explained by (1) seasonality and/or differences in the shelf characteristics, including the terrigenous DOM compositions, salinity ranges, and time scales of the water mass mixing, or (2) different approaches to evaluate the DOC behavior. It was also noted that the photodegradation of terrigenous DOM mainly occurs beyond the shelf, where the residence time of the water is much longer 44 . The application of the approach used in this study to evaluate the conservativeness of DOM in other shelf regions/basins and other seasons could be helpful to clarify the dynamics of DOM in the Arctic Ocean.

Methods
Field samples and measurements were collected during the T/S Oshoro-Maru C255 cruise, conducted from the southern Chukchi Sea to the Bering Strait in the early summer (July 9-20, 2013) (Fig. 1a). In 2013, the sea ice retreated from the Bering Strait to the north during June, but covered the adjacent to the northernmost stations in mid-July (Fig. 1a). Eighty-three surface water (1-3 m depth) samples were collected using a towed fish metal-free sampling system 51,52 . Surface water continuously flowed from the top of the towed fish to the onboard laboratory through a Teflon tube by an air-driven Teflon pump (model PFD-2, Asti co.). The temperature was monitored with a conductivity-temperature sensor (SBE 45, SeaBird ltd.). The time of the water flow from the towed fish to the sensor was less than one minute. Samples for salinity and nutrient analyses were collected without filtration. Samples for the DOC and DOM optical property analyses were filtered through an acid-cleaned 0.22-μ m filter unit (Millipak-100, Millipore) at the end of the tubing. The filtrate was collected in pre-combusted borosilicate glass vials and stored immediately at − 20 °C until analysis, within four months for optical analyses.
Two sea ice blocks (floes) were collected near the ice edge of the southern Chukchi Sea, using a nylon sling mounted on a vinyl coated stainless frame, which was hung from the ship's crane. A block of sea ice was cut into small pieces using a ceramic knife. A piece of the sea ice was put into an acid-cleaned HDPE bucket and melted in a dark at room temperature. Immediately after melting, the melted sea ice was filtered through an acid-cleaned 0.22 μ m filter (Durapore, Millipore) under a gentle vacuum, collected in pre-combusted borosilicate glass vials, and stored immediately at − 20 °C until analysis.
The salinity and nutrient concentrations were measured using a salinometer (AUTOSAL8400B, Guildline Instruments) and an autoanalyzer (QuAAtro2-HR, BL-Tec), respectively. N*, a potential tracer of denitrification, was calculated according to ref. 53. The DOC analysis was conducted by high-temperature combustion using a total organic carbon analyzer (TOC-V CSH , Shimadzu). The accuracy and consistency of the measured DOC concentrations were checked by analyzing a deep seawater reference standard (CRM program, Dr. Hansell Lab., University of Miami). After the water sample was thawed and reached room temperature, the absorbance of the sample was measured from λ = 200 nm to 800 nm at 0.5 nm intervals using a spectrophotometer (UV-1800, Shimadzu), according to ref. 54. A 5-cm quartz-windowed cell was used for the analysis. The spectral slope coefficient, between 275 nm and 295 nm (S 275-295 ), was calculated according to ref. 30. A smaller value of S 275-295 indicated a greater contribution of terrigenous DOM and vice versa 1 . The specific UV absorbance (SUVA 254 ), an indicator of the DOM aromaticity, was determined by dividing the absorbance measured at 254 nm by the DOC concentration 31 .
Excitation-emission matrix (EEM) fluorescence was measured using a fluorometer (Fluoromax-4, Horiba), according to ref. 55. The inner filter effect was corrected using the absorbance spectrum, according to ref. 56. Fluorescence intensities were corrected for the area under the water Raman peak (excitation = 350 nm), analyzed daily, and were converted to Raman units 57 . Parallel factor analysis (PARAFAC) was performed in MATLAB (Mathworks, Natick, MA) with the DOMFluor toolbox (version 1.7) 58 . The EEMs of the excitation wavelengths from 250 nm to 450 nm and emission wavelengths from 320 nm to 520 nm were used for PARAFAC modeling, and the three-component model was validated by split half validation and random initialization 58 . The relative Scientific RepoRts | 6:34123 | DOI: 10.1038/srep34123 abundance of each of the three fluorescent components (%Ci, i = 1 to 3) was calculated using the fluorescence intensity of individual components (C1, C2, and C3), i.e., %Ci = Ci / (C1 + C2 + C3) × 100.