Study area

Samples were collected from pack ice in the southern Sea of Okhotsk during two cruises on board the Japanese Coast Guard vessel Soya on 11–13 February 2008 and 9–12 February 2009, from landfast ice during land-based sampling in Saroma-ko lagoon (hereafter Saroma) on 3–4 March 2008 and 20 February 2009 (Fig. 1, Saroma), and from nearshore drift ice during land-based sampling off the Utoro Peninsula in the southeastern Sea of Okhotsk on 7 March 2008 (Fig. 1, site U). At each location sea ice and under-ice water (1-3 m from the ice underside) were collected. In total 14 sites were sampled (Table 1), with 14 sea-water samples and 51 ice sections (see below).

Table 1. Sampling sites and conditions

Fig. 1. Location of sampling sites in 2008 (squares; station number below symbol) and 2009 (circles; station number above symbol) in the southern Sea of Okhotsk. Saroma-ko lagoon (Saroma) and Utoro (U) sampling sites are also indicated.

Field sampling and sample processing

Sea-ice samples were collected using an ice corer (Mark II, Kovacs Enterprises, Inc., USA). One core was used to measure ice temperature (Testo 110 NTC, Brandt Instruments, Inc., USA); a second core was used for the analyses. In 2009, pancake and nilas ice (Table 1) were sampled from a basket lowered from the ship to the sea surface. One half of the core or part of the collected ice was used for immediate processing and subsampling. The other half of the core was placed in plastic tubing or bags and stored in an on-board freezer at -15°C.

For pigment analysis, ice samples were cut into sections, usually every 5 or 10cm, and melted directly on board the ship with the addition of artificial sea water (salinity 40). Samples were melted at room temperature in the dark in acid-washed lidded Teflon cups.

The second half of the ice-core samples from the Soya expeditions were placed in polyethylene tubing immediately after collection and kept horizontal in a freezer (at - 15°C) during the cruise. These half-core samples from Soya for CDOM, salinity and δ¹⁸O analyses were processed in the freezer room at Hokkaido University shortly after each cruise. The core half was cut into a 4 cm x 3 cm rectangular cross section and then cut into sections. To avoid contamination during sampling and handling processes, 3 mm of the outside of the ice sections were removed with a clean stainless-steel plane. The remaining ice was placed in an acid-washed Teflon cup for melting at room temperature in the dark.

Sea-ice samples from land-based sampling in Saroma and Utoro were processed directly on site (pigment analysis). Immediately after retrieval the samples were cut into 5-10cm sections and placed in pre-cleaned lidded polyethylene cups for melting. Samples for pigment analysis were melted as for the samples from Soya. Frozen cores were taken to Hokkaido University in Sapporo and processed for salinity, δ¹⁸O and CDOM analyses as described for the samples from Soya.

Under-ice water samples were collected on all occasions through the ice-core holes using a 500 mL Teflon water sampler (GL Science Inc., Japan), usually at 3 m depth (Table 1), ∼15min after drilling of ice cores to avoid disturbance caused by drilling.

Melted sea ice and under-ice water were subsampled for salinity and oxygen isotopic ratio (δ¹⁸O) measurements into 10mL glass vials, and 40 mL pre-combusted amber glass vials for CDOM. Salinity and δ¹⁸O samples were kept at room temperature (+15°C). CDOM samples were filtered through PALL Acrodisk^® membrane filters (Supor^® membrane, pore size 0.2 m, with 0.8m Supor^® pre-filter) using acid-washed all-plastic syringes. Ultrapure water treated as samples showed no trace of leaching of CDOM into the samples during melting, filtering and storage. CDOM samples were stored in the dark at +4°C. Samples of known volume were filtered onto Whatman GF/F glassfiber filters (nominal pore size 0.7 m) using low vacuum for pigment analysis. Filters were placed in liquid nitrogen until measurements were conducted.

Analytical methods

Salinity was measured using a salt analyzer (SAT-210; Toa Electronics Ltd, Japan). A standard deviation for salinity calculated from 15 subsamples taken from a reference water sample (S= 10.00) was 0.03. δ¹⁸O was determined with a mass spectrometer (DELTA plus; Finnigan MAT, USA). Results are reported as δ¹⁸O (‰), defined as the deviation of H₂ ¹⁸O/H₂ ¹⁶O ratio of the measured sample to that of Vienna Standard Mean Ocean Water (VSMOW). The precision of δ¹⁸O analysis from duplicate determinations is ±0.02% (Reference Toyota, Takatsuji, Tateyama, Naoki and OhshimaToyota and others, 2007). Chlorophyll a (Chl a) was determined with high-performance liquid chromatography (HPLC; Shimadzu SPD-M10AVvp with Class-VP Multi Workstation) following the methods of Reference Furuya, Hayashi and YabushitaFuruya and others (1998).

CDOM spectral absorbance (A(λ)) was measured using a Shimadzu UV-PC2501 spectrophotometer with a 10 cm quartz cell, from 190 to 800 nm, with 2 nm slits and at 1 nm intervals, after samples had been allowed to warm to room temperature, with ultrapure water (Milli-Q) as reference. A baseline correction was carried out, with average absorbance at 690-710 nm subtracted from the spectra. Absorbance was converted to absorption coefficients (a _CDOM(λ) (m^-1) = [2.303∆(λ)]/0.1 m). CDOM samples were measured within 2-3 months after collection. The spectral slope coefficient S at 250-400 nm range was calculated using a linear fit to log-transformed absorption data.

The CDOM fluorescence was measured using a Varian Cary spectrofluorometer with excitation at 240-450 nm (every 5 nm), and the resulting emission was measured from 300 to 600 nm (every 2 nm). The resulting excitation-emission matrices (EEMs) were corrected for instrument-specific spectral corrections, inner filter effects (based on the absorption of the sample) and Raman and Rayleigh scattering, and finally converted to Raman units (RU) based on the Raman signal from ultrapure water (ultrapure water was run as a sample for each batch of samples measured to account for possible instrument drift). Corrected fluorescence data (i.e. EEMs) were analyzed using the DOMFluor Toolbox v1.7 (Reference Stedmon and BroStedmon and Bro, 2008).

PARAFAC modeling

A total of 65 EEMs were collected and modeled in this study, which included 14 sea-water samples and 51 samples from sea-ice cores. However, the dataset was very diverse; it was not collected along a mixing gradient or time series, but contained extreme samples. Therefore it is not ideal for parallel factor analysis (PARAFAC) modeling (see Reference Stedmon and BroStedmon and Bro, 2008). Thus weighing was carried out in accordance with Reference Kowalczuk, Durako, Young, Kahn, Cooper and GonsiorKowalczuk and others (2009), who also had a diverse EEM dataset. Samples with extremely high fluorescence were diluted, such that all samples were within a maximum intensity range of 0-0.5 RU. This weighed dataset was split-half validated for a five-component PARAFAC model, and a residual analysis also confirmed that a five-component model was appropriate (Reference Stedmon and BroStedmon and Bro, 2008). Table 2 lists the excitation and emission peaks of the components derived for the validated PARAFAC model. The fluorescence intensities in the subsequent data analysis refer to the maximum fluorescence signals of each component.

Table 2. Primary and secondary excitation–emission (Ex/Em) peaks for PARAFAC model components on Okhotsk sea ice and surface waters. Median EFs for components in sea ice relative to underlying sea water

Sea-water and sea-ice characteristics

In 2008 the ice sampled was first-year ice (˃0.3 m thick), while in 2009 the majority of the ice sampled in the Sea of Okhotsk was young ice, pancake ice or dark nilas (˂0.3 m thick). This is reflected in the higher salinity of the younger ice (Table 1). From previous studies it is known that snow can contribute significantly to sea-ice mass both in the southern Sea of Okhotsk (Reference Ukita, Kawamura, Tanaka, Toyota and WakatsuchiUkita and others, 2000; Reference Toyota, Takatsuji, Tateyama, Naoki and OhshimaToyota and others, 2007; Reference NomuraNomura and others, 2010) and in Saroma (Reference Shirasawa, Leppäranta, Saloranta, Kawamura, Polomoshnov and SurkovShirasawa and others, 2005; Reference Nomura, McMinn, Hattori, Aoki and FukuchiNomura and others, 2011), although considerably more so in the latter. Our observations concurred with earlier observations: in 2008 -10% of the total length of the Okhotsk ice samples was snow-ice, while in Saroma snow-ice contributed 30-40% of the total ice thickness based on δ¹⁸O composition. Dynamic thickening and rafting of young floes contributed significantly in the Sea of Okhotsk in 2008. In 2009, as mainly new or young ice was sampled, the snow contribution to ice mass was negligible in the Okhotsk samples. In Saroma, snow-ice again contributed significantly, although less than in 2008.

Reference Uusikivi, Vähätalo, Granskog and SommarugaYamamoto and others (2001, 2002) proposed that deviations in sea-water composition (salinity and δ¹⁸O) from the western sub-arctic Pacific water (WSAP) line (δ¹⁸O = 0.3915x salinity-13.561, R ² = 0.99, n = 59; Reference Yamamoto, Tanaka and TsunogaiYamamoto and others, 2001) can be used to explain the mechanism behind the sea-water composition. Our data are mainly displaced in a manner consistent with the addition of meteoric water (lighter δ¹⁸O) and brine from sea-ice formation (higher salinity), relative to the WSAP line (not shown) in both Saroma and Okhotsk samples. Samples collected 3 m below the ice bottom (2008) might not capture the shallow fresher under-ice layer in Saroma (see Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009), which is reflected in the lower salinity (Table 1) and light isotope values for under-ice water collected 1 m below the ice bottom in Saroma in 2009 (not shown).

Algal biomass

The algal biomass in sea ice in Saroma ranged from 2.0 to 9.0mgChl am–² in 2008 and 0.3 to 10.6mgChl am–² in 2009. This contrasts with previous observations which include 272.8±20.2 mgChl am–² (Reference McMinn and HattoriMcMinn and Hattori, 2006), 2-119 mg Chl am–² (Reference Robineau, Legendre, Kishino and KudohRobineau and others, 1997) and 38.25 mgChl am–² (Reference Kudoh, Robineau, Suzuki, Fujiyoshi and TakahashiKudoh and others, 1997). Although biomass in Saroma was thus lower than in earlier studies, biomass in Okhotsk ice was even lower than in Saroma, where it ranged from 0.2 to 1.6 mgChlam^–2 in 2008 and from 0.0 to 3.5 mg Chl a m–² in 2009. The ice algal biomass distribution was typically accumulated in the bottom 5-10 cm. In only one core (2008) from Saroma was algal biomass distributed evenly throughout the length of the core (Saroma R station). Chl a levels were up to an order of magnitude lower in under-ice water than in ice, both in Okhotsk and Saroma. The standing stock in the ice cores corresponded to the total algal biomass in 5-15 m of the under-ice water column at the time of sampling.

CDOM absorption

Measurements of CDOM absorption in sea water showed the typical featureless exponential decay (e.g. Reference Bricaud, Morel and PrieurBricaud and others, 1981) with increasing wavelength, as shown in Figure 2a and b. Values of absorption at 350 or 375 nm (a₃₅₀ and a₃₇₅) were higher in Saroma than in Okhotsk waters (Table 3). The spectral slope coefficient S had values between 0.0197 and 0.0280 in Okhotsk and 0.0183 and 0.0202 in Saroma (Table 3). The lower S values in Saroma imply a greater influence from rivers (e.g. Blough and Del Vecchio, 2002), and the lower values in 2009 vs 2008 also exemplify the less saline water sampled in 2009 belonging to a well-developed under-ice plume (Reference Nomura, Takatsuka, Ishikawa, Kawamura, Shirasawa and Yoshikawa-InoueNomura and others, 2009). The majority of the sea-ice samples had local absorption peaks (Fig. 2c and d), which made it impossible to compute S values as the fit was very poor. A distinct peak at -325 nm was present in the majority of sea-ice samples, while a secondary peak was visible in some sea-ice samples at -275 nm. Peaks in CDOM absorbance spectra similar to those observed here were reported by Reference Belzile, Johannessen, Gosselin, Demers and MillerBelzile and others (2000) in ice-core samples from Baffin Bay, Canadian Arctic, and similar peaks have recently been found to be a ubiquitous feature of CDOM in Antarctic sea ice (Reference NormanNorman and others, 2011). Patasayeva and others (2004) observed absorption peaks between 270 and 300 nm in sea-ice brine samples generated in an ice tank experiment. Similar absorbance spectra have also been found in the CDOM produced by cyanobacteria in culture (Reference Steinberg, Nelson and PrusakSteinberg and others, 2004). The exact source of these peaks is unknown; however, they seem to represent an ice-specific CDOM fraction, i.e. related to autochthonous production, and are suggested to be related to the production of ultraviolet-absorbing compounds for photoprotection (e.g. Reference Belzile, Johannessen, Gosselin, Demers and MillerBelzile and others, 2000). It is known that mycosporine-like amino acids (MAAs) absorb strongly at -330nm, while at 275-280nm aromatic amino acids (precursors to MAAs) and some nucleic acids absorb strongly (e.g. Reference Woźniak and DeraWoźniak and Dera, 2007). All these compounds are produced by marine organisms. MAAs were almost completely absent at the bottom of thick Antarctic sea ice (Reference Ryan, McMinn, Mitchell and TrenerryRyan and others, 2002), likely due to low exposure to ultraviolet light. In thin (snow-free) seasonal sea ice in the Baltic Sea, more representative of the conditions in the Sea of Okhotsk or Saroma, concentrations of MAAs were exceptionally high (with MAA to Chl a ratios up to 0.65) and could be detected as high particulate absorption peaks at -330 nm (Reference Uusikivi, Vähätalo, Granskog and SommarugaUusikivi and others, 2010).

Table 3. Optical properties of CDOM in sea water and sea ice. Mean and standard deviation (maximum) are given for a₃₅₀, a₃₇₅ and a_R (ratio a₃₃₀: a ₃₀₀). Range is given for S. S values for sea ice are not given as they could not be computed reliably

Fig. 2. Absorption spectra for sea-water (a, b) and sea-ice (c, d) samples collected in 2008 and 2009, respectively. Blue and green lines in (a–d) show the mean. The means in 2008 (e) and 2009 (f) for sea water (blue) and sea ice (green) are compared. Note the variable y-axis scale.

From Figure 2c and d it is apparent that many of the absorption spectra in sea ice are often very flat below 350 nm and do not have the characteristic strong exponential increase in CDOM absorption at shorter wavelengths (e.g. Blough and Del Vecchio, 2002). The samples with these flat absorbance spectra are predominantly from older first-year sea ice, while the younger ice tended to have absorbance spectra similar to that of sea water (besides the distinct absorption peaks also present in some of these samples). This could be partly due to the age of the ice, since either mature sea-ice communities produced this material or sunlight photobleached dissolved material, thereby flattening the absorption spectra. Flatter CDOM absorption spectra at shorter wavelengths have been observed in surface ice layers and for older ice in the Baltic Sea (Reference Uusikivi, Vähätalo, Granskog and SommarugaUusikivi and others, 2010; Muller and others, 2011).

Our observations indicate that sea ice can support concentrated production of specific light-absorbing compounds, presently of unknown origin or composition, which can be released during ice melt. Values in sea ice are considerably higher than expected from conservative incorporation relative to salinity. The absorption peak at -330 nm was particularly distinct in the majority of the sea-ice samples, while the absorption at -300nm has values expected from simple exponential decay with increasing wavelength. Hence, the ratio of absorption at 330 nm to that at 300 nm (a_R = a₃30: a₃₀₀), which was considerably higher in sea ice than in sea water (Table 3), can be used to show that CDOM released from sea ice appears to differ substantially from that of underlying sea water, although the concentration appears similar to sea water. Thus, sea ice does not provide an excess amount of CDOM during melt, as reported for Baffin Bay (Reference Nakatsuka, Toda, Kawamura and WakatsuchiScully and Miller, 2000), but rather is a source of CDOM of different quality (Reference NormanNorman and others, 2011).

CDOM fluorescence

Characteristic fluorescence excitation–emission

The majority of sea-water samples were dominated by a terrestrial humic-like fluorescence (Fig. 3) with excitation at 250 nm and maximum emission at ˃400nm (Reference CobleCoble, 1996). Some sea-water samples had another fluorescent compound with excitation/emission maximum at -275/305 nm. The EEMs in sea ice differed considerably from those in (parent) sea water (Fig. 3), and are dominated by emission below 400nm, similar to observations in Baltic Sea ice (Reference Stedmon, Thomas, Granskog, Kaartokallio, Papadimitriou and KuosaStedmon and others, 2007).

Fig. 3. Typical EEMs in the study region. EEMs shown are means for sea water in Saroma (a), sea water in the Sea of Okhotsk (b), sea ice in Saroma (c) and sea ice in the Sea of Okhotsk (d). Note the difference in intensity scale between sea-water (a, b) and sea-ice samples (c, d).

PARAFAC components

Components in the validated PARAFAC model (Table 2) overlap with results from earlier studies. Two of the identified components (C1 and C2) had properties that are often referred to as humic-like (Reference CobleCoble, 1996), with broader excitation spectra and emission maxima ˃380 nm. The other three components (C3-C5; Table 2) resemble protein-like material, with emission at ˂360nm (Reference CobleCoble, 1996), overlapping with findings from other studies. For example, C3 resembles tryptophan-like (free or bound to protein) fluorescence in Antarctic sea-ice brines (Reference Stedmon, Thomas and PapadimitriouStedmon and others, 2011), and C4 and C5 have been referred to as tyrosine-like fluorescence (Reference CobleCoble, 1996; Patasayeva and others, 2004). These are likely not pure compounds but bound to other compounds/molecules, which is why their fluorescence characteristics differ from those of pure amino acids.

Sea-water fluorescence in both Saroma and Okhotsk was dominated by the humic-like component C2, with a significant negative correlation with salinity and δ¹⁸O in sea water (r˂-0.76, p˂0.01), implying a terrigenous source. This relationship could not be found in sea ice. In contrast, sea-ice DOM fluorescence was generally dominated by varying degrees of the protein-like components. To examine whether these compounds are found in excess in sea ice relative to parent sea water, enrichment factors (EFs) were calculated as in Reference Müller, Vähätalo, Stedmon, Granskog and NormanMüller and others (2013), with values ˃1 indicative of an excess in sea ice that cannot be explained by conservative incorporation relative to salinity during ice formation (Table 2). The EFs for protein-like components are much higher (median values 8.9–14.0; Table 2) than observed during initial ice formation (EFs˂ 3; Reference Müller, Vähätalo, Stedmon, Granskog and NormanMüller and others, 2013, Fig. 11). EFs for humic-like material (EF ˂ 4) were considerably lower than for protein-like material, similar to what is found during initial ice formation (Reference Müller, Vähätalo, Stedmon, Granskog and NormanMüller and others, 2013). The considerable protein-like accumulation thus points towards biological activity in sea ice during ageing (cf. Reference Stedmon, Thomas, Granskog, Kaartokallio, Papadimitriou and KuosaStedmon and others, 2007, 2011).

The much higher biomass in sea ice relative to sea water points towards possible accumulation in sea ice as a result of sympagic production. CDOM released during sea-ice melt is therefore compositionally different than that present in parent sea water, and sea ice might therefore be a source of fresh and potentially more labile CDOM (Reference NormanNorman and others, 2011). However, more detailed studies during ice melt are needed to validate this assumption. Reference NormanNorman and others (2011) showed that the material released from sea ice is susceptible to exposure to sunlight and is easily photobleached; therefore, its lifetime could be transient.