Seasonal phototransformation of dissolved organic matter to ammonium, dissolved inorganic carbon, and labile substrates supporting bacterial biomass across the Baltic Sea

Authors


Abstract

[1] We studied the photochemical transformation of dissolved organic matter (DOM) to dissolved inorganic carbon (DIC), ammonium (NH4+), and labile organic substrates supporting bacterial carbon biomass along a salinity gradient throughout the Baltic Sea during summer, autumn, and spring. The photoproduced DIC, NH4+, and labile DOM supporting bacterial biomass were related to the number of photons absorbed during the irradiations of biologically recalcitrant DOM to determine apparent quantum yields. The apparent quantum yields for the photoproduction of DIC and NH4+ lacked seasonal variation, but behaved differently along the salinity gradient; the photoproduction of DIC decreased, while photoammonification increased with increasing salinity. The apparent quantum yield for the photoproduction of labile DOM supporting bacterial biomass was highest in summer and unaffected by salinity. The annual photoammonification rate over the entire Baltic Sea ranged from 0.038 to 0.049 Tg N, equivalent to 13%–23% of the annual atmospheric deposition of inorganic N. The annual phototransformation of dissolved organic carbon (DOC), including the direct photomineralization and indirect bacterial mineralization of photoproduced labile DOM (total of 2.71–3.94 Tg C), exceeded the annual river loading of photoreactive DOC, assuming that half of the total river DOC input to the Baltic Sea is photoreactive. As the annual photomineralization of DOC exceeded the annual terrestrial input of photoreactive DOC to the Baltic Sea, the photochemical transformation is a major sink for terrestrial DOC in such coastal systems.

1. Introduction

[2] Dissolved organic matter (DOM) constitutes a large reservoir of carbon (C) and nitrogen (N) in aquatic ecosystems. The sources of DOM in coastal waters include primary production (autochthonous sources) and riverine input of terrestrial DOM from the surrounding catchment (allochthonous sources). Allochthonous DOM, including humic substances, is mainly biologically recalcitrant, but solar radiation may photochemically decompose such DOM in surface waters. In coastal waters with large inputs of allochthonous DOM, photochemistry forms a potential sink for terrestrial organic matter [Miller and Zepp, 1995; Hedges et al., 1997].

[3] Solar radiation can transform biologically recalcitrant DOM into labile constituents and support bacterial growth [Moran and Zepp, 1997; Miller et al., 2002; Obernosterer and Benner, 2004]. Such phototransformation of DOM increases both bacterial respiration and biomass. The latter transfers biologically recalcitrant but photochemically reactive DOM to higher trophic levels [Vähätalo et al., 2011].

[4] Solar radiation also mineralizes dissolved organic carbon (DOC) directly to carbon dioxide (CO2, counted as part of dissolved inorganic carbon, DIC) or to carbon monoxide (CO) [Mopper et al., 1991; Mopper and Kieber, 2000; White et al., 2010], of which DIC is the dominant photoproduct [Miller and Zepp, 1995]. Solar radiation may transform dissolved organic nitrogen (DON) to ammonium (NH4+), which enhances the growth of bacteria and algae in nitrogen-limited surface waters [Bushaw et al., 1996; Morell and Corredor, 2001; Vähätalo and Zepp, 2005; Vähätalo et al., 2011].

[5] Comparing photochemical reaction rates among studies is difficult without normalization to the amount of light absorbed by the water samples during irradiation and without taking into account experimental conditions such as temperature and the duration of irradiation. Photochemical reactions are typically faster at higher temperatures [Turro, 1991; Schwarzenbach et al., 2003; Zhang et al., 2006]. On the other hand, reaction rates slow as the photoreactivity of DOM decreases over longer irradiation periods [Vähätalo and Wetzel, 2004]. The decrease in photoreactivity may be linked to the photochemical bleaching of chromophoric dissolved organic matter (CDOM) [Vähätalo and Wetzel, 2004] and may explain the decrease in the photomineralization of DOC observed during a summer in a small humic lake [Vähätalo et al., 2003]. Such seasonal differences in the photoreactivity of DOM have not been examined in coastal waters.

[6] The importance of the photomineralization of DOM in carbon cycling has been examined in a few lakes [Molot and Dillon, 1997; Pers et al., 2001; Kopácek et al., 2004]. Annual photomineralization contributes 10% to the total removal of organic carbon in Pleŝné Lake and Lake Örträsket [Pers et al., 2001; Kopácek et al., 2004], while in a few low-DOC lakes in Canada, photomineralization dominates the total losses of allochthonous DOC [Molot and Dillon, 1997]. Photomineralization is a major sink for terrestrial DOC in coastal waters also based on estimates of the photoproduction of labile organic carbon or the photochemical consumption of oxygen assessed for terrestrial DOC in coastal waters on a global scale [Miller and Zepp, 1995; Andrews et al., 2000; Miller et al., 2002]. These estimates are based on few measurements in the laboratory and rough extrapolations over the global coastal ocean. In contrast, a recent study estimated that <7% of Mackenzie River DOC will be photochemically mineralized in a coastal section of the Arctic Ocean [Bélanger et al., 2006]. Obviously, more detailed studies are needed to assess the photochemical decomposition of DOM in mid- and low-latitude coastal waters.

[7] The quantitative assessment of photoreactions in coastal waters is challenging. For example, the delineation of coastal waters is unclear, and the quantification of water exchange between open and coastal ocean is difficult. A further challenge in coastal waters is the influence of salinity on the photomineralization of DOC, which has been reported to decrease with increasing salinity [Bélanger et al., 2006; White et al., 2010]. At mid-high latitudes, seasonal changes in solar irradiance [Zepp and Cline, 1977] and the temperature of surface waters [Zhang et al., 2006] also affect photochemical rates. The suggested importance of photoreactions (including direct and indirect mineralization) as a sink of terrestrial DOC therefore requires further quantitative verification that accounts for the major factors regulating the photochemical reaction rates of DOM in coastal waters.

[8] The aim of this work was to study the direct photochemical mineralization of DOC and DON, and the photoproduction of labile DOM supporting bacterial biomass from the same water samples along a salinity gradient in the Baltic Sea during spring, summer, and autumn. We quantified the photochemical reactivities (i.e., apparent quantum yields) for the photoproduction of DIC, NH4+, and labile DOM supporting bacterial biomass using simulated solar radiation, and estimated the seasonal and annual rates for these photoreactions over the entire Baltic Sea. The importance of annual photoreaction rates was assessed by comparing the photoreactions to other C and N fluxes in the Baltic Sea. We found that the annual phototransformation of DOM corresponded to 9%–18% of the annual river loading of DON, but exceeded the annual river loading of photoreactive DOC into the Baltic Sea.

2. Materials and Methods

2.1. Research Site and Sampling

[9] The Baltic Sea is a semi-enclosed shallow sea (mean depth 54 m) with a drainage area 4.2 times as large as the sea area (380 000 km2; Figure 1). The water renewal time is 50 yrs [Leppäranta and Myrberg, 2009]. The large riverine inputs of freshwater (450 km3 yr−1) and allochthonous material result in a salinity of 0–8 in the surface water and high concentrations of DOC and DON in the Baltic Sea [Elmgren, 1984; Leppäranta and Myrberg, 2009].

Figure 1.

The sampling stations and the isohalines in the Baltic Sea according to Voipio [1981]. The stations are (I) Arkona Sea, (II) Gotland Basin, (III) Gulf of Finland, (IV) Helsinki, (V) Neva Bay, and (PJ) Lake Pääjärvi.

[10] In July and September 2006, and March 2007, surface water samples were collected automatically from the flow-through system of ship (from approximate depth of 4 m) with a refrigerated sequence sampler (Isco 3700 R) from Helsinki to the Arkona Sea [Leppänen and Rantajärvi, 1995]. Water samples from the estuary of the Neva River, the largest freshwater source of the Baltic Sea (Figure 1), were collected with a Limnos sampler from 0 to 3.5 m (at maximum depth of 4 m). The two sampling methods are comparable, as the collected water samples represent the surface waters in the Baltic Sea at the moment of collection. The water samples were kept in the dark for 2–3 days before arriving at our laboratory. In order to reduce the concentration of biologically labile dissolved organic matter, GF/F-filtered (Whatman, nominal pore size ca. 0.7 μm) water samples were incubated at 22°C–24°C in the dark for 4–37 d prior to photochemical experiments. All glassware was rinsed with detergent, 6% HCl, and ultraclean water (Millipore Milli-Q), and precombusted (2 h at 450°C) before use.

2.2. Nutrient and Carbon Analysis

[11] The concentrations of nitrate (NO3) and nitrite (NO2) were measured as triplicates according to Grasshoff et al. [1983, 1999], and ammonium (NH4+) as 5 replicates with 5 measurements from each of them with a phenol-hypochlorite method [Solórzano, 1969]. Dissolved organic carbon (DOC) and total dissolved nitrogen (TDN) were measured as three replicates from acidified samples (pH lowered to 2 with 2 mol l−1 HCl) with a high temperature catalytic oxidation method using a carbon and nitrogen analyzer (Shimadzu TOC-V CPH; the coefficient of variation (CV) of the method was 5% for TDN and 7% for DOC). The concentrations of dissolved organic nitrogen (DON) were calculated by subtracting the measured NO3, NO2, and NH4+ concentrations from the TDN concentrations.

2.3. Photochemical Experiments

[12] The water samples were filtered (0.2 μm Pall Supor 200) and sealed without a headspace in 170 ml quartz bottles (as samples to be irradiated) or in borosilicate bottles wrapped in aluminum foil (as dark samples) or kept in the dark in ice water (as initial samples). The samples were irradiated for 46–71 h with 765 W m−2 simulated solar radiation (Atlas Suntest CPS+ Solar simulator) [Vähätalo and Zepp, 2005]. The quartz bottles were kept on their side above the dark samples and 3 cm below the surface of flow-through cooling water regulated at 5°C (Millipore Milli-ρ regulated with a Lauda RE 112 thermostat). The low irradiation temperature was selected to minimize microbial interference. The bacterial densities in <0.2 μm-filtered waters were negligible (typically <3 × 103 cells ml−1) throughout experiments (methods explained below).

2.3.1. Photoammonification

[13] After the irradiation, we measured the median concentrations of NH4+ in the irradiated, the dark, and the initial samples (5 measurements from 5 bottles of each treatment). To measure the photoammonification with a high precision but without a temporal drift, a dark sample was always measured before its corresponding irradiated sample. The difference between the irradiated and the dark sample described the amount of photoammonification.

2.3.2. Photoproduction of Dissolved Inorganic Carbon (DIC)

[14] We determined the photomineralization of DOC as the photoproduction of dissolved inorganic carbon (DIC). Before the irradiations, the indigenous DIC was removed from the filtered (0.2 μm Sartobran 300) samples. The samples were acidified to pH 4 with 1 mol l−1 HCl, and the resultant CO2 was bubbled away with CO2-free air. Prior to the irradiations, pH was adjusted back to the ambient level (8–8.5) with 0.1 mol l−1 NaOH [Johannessen and Miller, 2001; Bélanger et al., 2006; White et al., 2010]. After the irradiations, the DIC concentrations were measured from the initial, the dark, and the irradiated samples (in this specific order to maximize the precision of the measured difference between the irradiated and the dark samples) by injecting a 750 μl water sample (7 injections per sample) into an acid bath and purging the resultant CO2 to a CO2 detector [Salonen, 1981] (CV < 1%). The photochemical mineralization of DOC was calculated as the difference in the DIC concentrations between the irradiated and the dark samples.

2.3.3. Photoproduction of Labile DOM Supporting Bacterial Biomass (bact-C)

[15] To measure the production of bacterial biomass based on the photoproduced labile DOM, we incubated the 0.2 μm-filtered (Pall Supor 200) irradiated and dark samples with indigenous bacteria from each sample in the dark for 7–12 d at 22°C–24°C. The GF/F (Whatman) filtrate from initial water sample was added as final concentration of 10% of the total water volume (140 ml). To avoid any nutrient limitation in bacterial growth, the samples received nutrients (7.1 μmol NH4Cl l−1 and 1.4 μmol KH2PO4 l−1). Along the incubation, one 10 ml subsample was preserved daily with glutaraldehyde (final concentration, 5%), filtered on Poretics polycarbonate 0.22 μm filters and stained with acridine orange [Hobbie et al., 1977]. To calculate the bacterial biomass, we counted ≥200 bacterial cells from ≥20 fields with an epifluorescence microscope (Leitz Aristoplan) and determined the cell volumes using digital image analysis [Massana et al., 1997]. The biovolumes were converted to bacterial carbon through a conversion factor of 0.12 pg C (cell volume μm3)0.7 [Norland, 1993]. When the highest bacterial cell densities were reached in the irradiated samples, the photoproduction of labile DOM supporting bacterial carbon biomass was calculated as the difference in bacterial biomass between the irradiated and the corresponding dark sample. The coefficient of variation for bacterial biomass (CVBB) was calculated as a combined variety of bacterial cell densities and cell volumes, when CVBB = (CVD2 + CVV2 + 2CVD CVV)1/2, where CVD is the CV for bacterial density and CVV is the CV for bacterial cell volumes [Wilkinson, 1961].

2.4. The Number of Photons Absorbed by the Irradiated Samples

[16] The number of photons absorbed by the samples was calculated by accounting for the spectral doses of photons, the optical path length, and the absorption characteristics of samples (equations (S1) and (S2) in the auxiliary material (Text S1)). The measurements of chromophoric dissolved organic matter (CDOM) are described below, and further details are provided in the auxiliary material. The absorption by CDOM was measured after every irradiation experiment from the irradiated, the dark, and the initial samples with a spectrophotometer (Shimadzu UV-2100) using a 10 cm quartz cuvette. The absorbance spectrum of samples and blanks (Millipore Milli-Q water) was scanned three times between 190 and 800 nm in 1 nm steps against an empty reference cell holder. The blank was subtracted from the sample and the resulting absorbance of CDOM (ACDOM,λ) was converted to an absorption coefficient (aCDOM,λ) by 2.303ACDOM,λ 0.1−1. The spectral slope coefficient (S) of aCDOM,λ was calculated by the ln-linear-fitting method to a wavelength range from 300 nm to 550 nm, which was the wavelength having the detection limit (>6 × SD) of our method. For wavelengths >550 nm, aCDOM,λ was estimated by applying the measured S to aCDOM,300 exp[−S(λ − 300 nm)].

2.5. Models for Photodecomposition of DOM

[17] The photoreaction rate at depth z (prz; mol product m−3 d−1) is a product of three spectra [details in Vähätalo et al., 2000]:

equation image

where ϕλ is the apparent quantum yield spectrum for the photoreaction (mol product (mol absorbed photons)−1 nm−1), Qs,z is the scalar photon flux density spectrum at depth z (mol photons m−2 d−1 nm−1), and aCDOM,λ is the absorption coefficient spectrum of CDOM (m−1 nm−1).

[18] The photoreaction rate (pr; mol product m−2 d−1) in the entire water column is:

equation image

where Qλ represents the spectrum of photons absorbed by the entire water column (mol photons m−2 d−1 nm−1), and the ratio aCDOM atot−1(dimensionless) is the contribution of CDOM to the total absorption coefficient atot (m−1 nm−1). If CDOM absorbs all photons, the ratio aCDOM atot−1 is 1.

[19] The apparent quantum yield spectrum (ϕλ) for the photoreaction describes the amount of photoproduct (mol NH4+, DIC, or bact-C) per mole of photons absorbed [Hu et al., 2002]. On the basis of our laboratory irradiations, we calculated ϕλ from equation (1) by dividing the amount of photoproduct (prz) by the number of photons absorbed by the sample (Qabs,λ), which is product of Qs,z,λ and aCDOM. ϕλ was assumed to depend on wavelength according to:

equation image

where c (dimensionless) and d (nm−1) are positive constants, and λ is the wavelength (nm). The parameters c and d in equation (3) were iterated by relating the amount of photoproduct (mol NH4+, DIC, or C bound in bacterial biomass) to Qabs,λ (see auxiliary material) through unconstrained nonlinear optimization (the ‘fminsearch’ function of the Matlab 7.9.0), as in our earlier publications [Vähätalo et al., 2000; 2003; 2011; Vähätalo and Wetzel, 2004; Vähätalo and Zepp, 2005; Vähätalo and Järvinen, 2007]. The details of these calculations are given in the auxiliary material.

[20] The coefficient of variation (CV) for ϕλ was calculated as a combined variability of the parameters prz, Qs,z,λ and aCDOM,λ (equation (1)) used for the determination of ϕλ. The CV for Qs,z,λ among different samples irradiated with a Suntest CPS+ is <1% [Lam et al., 2003] and ignored here. Thus, CV% for ϕλ was calculated as CV = (CVpr,z2 + CVaCDOM,λ2 − 2CVpr,z CVaCDOM,λ)1/2, where CVpr,z and CVaCDOM,λ are the coefficients of variation for determination of the amount of the photoproduct (NH4+ or DIC) and the absorption coefficient of CDOM, respectively [Wilkinson, 1961]. The CV for the ϕλ describing labile organic substrates stimulating bacterial carbon biomass was calculated in the same way.

2.6. Validation of the Modeled Photodecomposition

[21] For validation, we compared the modeled (equation (2)) and the measured (in situ) rates for the photomineralization of DOC in a mesohumic Finnish lake (Lake Pääjärvi). To measure the photomineralization of DOC in situ, the 0.2 μm-filtered (Sartobran 300) and DIC-free lake water sample prepared as described above was shared into 20 ml quartz (as samples to be irradiated) and glass tubes (as dark samples). The samples were irradiated with natural solar radiation in a floating incubation rack in the lake at depths between 0.02 and 0.2 m for 24 h [Vähätalo et al., 2000]. The Al-foil-wrapped dark samples were incubated in a black plastic bag sunk next to the rack. The photomineralization rate over the entire water column (pm; mol C m−2 d−1) was calculated by integrating the rates measured over the depths according to:

equation image

where pm0 is the rate for the photomineralization of DOC at a depth of 0 m (mmol C m−3 d−1), Kd,pm is the extinction coefficient for photomineralization (m−1), z is the depth (m), and the integration goes through a depth of 0 m (zmin) to an infinite depth (zmax).

[22] In Lake Pääjärvi, the photomineralization of DOC fell steeply with water depth (Figure 2). The in situ measured rates integrated over the entire water column were higher than those modeled (Table 1) according to equation (4) and using the measured doses of photons (Qv,d,λ), the ϕDIC determined for Lake Pääjärvi water (with an irradiation time of 18.7 h), and an aCDOM,λ atot,λ−1 of 1. The Qv,d,λ was derived from the measurements with a spectroradiometer (Macam SR991) on the roof of the Lammi Biological Station (300 m from Lake Pääjärvi) at 30 min intervals over the time of the in situ incubations and accounting for the albedo of the lake (6%). The ϕDIC,λ was determined in the laboratory at a temperature of 5°C, which was lower than that of the surface water during the two in situ measurements (Table 1). We assumed that the difference between the measured and the modeled rates depends on the temperature difference between the laboratory (5°C) and the in situ (18.8°C or 16.9°C) irradiations. We assessed the temperature dependence of the photomineralization of DOC over the photolytic layer (0–1 m) based on the measured and modeled rates by determining the activation energy (Ea; J mol−1) of the Arrhenius equation:

equation image

where k is the decomposition coefficient for the photomineralization of DOC measured at the in situ temperature (Table 1) or calculated based on modeling with ϕDIC,λ determined at 5°C (d−1), A is the frequency factor (dimensionless), R is the general gas constant (J mol−1 K−1), and T is the temperature (K). The mean of two activation energy determinations for the photolysis of DOC to DIC was the 20.7 kJ mol−1 (Table 1), which is similar to 22.8 kJ mol−1 determined earlier for the photolysis of pyruvic acid to DIC [Guzmán et al., 2007]. When we used the mean Ea for the photomineralization of DOC (20.7 kJ mol−1) and accounted for the temperature difference between +5°C in the laboratory and that in situ, the difference between the modeled and measured rates was ±5% (Table 1). We previously obtained similar uncertainty between measured and modeled photoreactions rates for the photoproduction of DIC and the photobleaching of CDOM using the same modeling approach, but with ϕs determined at a temperature corresponding to measurements in situ [Vähätalo et al., 2000; Vähätalo and Wetzel, 2004]. Thus, the similarity of our Ea to the previous one [Guzmán et al., 2007] indicates that the modeling method applied here describes the rates of photochemical reactions well when temperature dependence on photoreactions is taken into account.

Figure 2.

The photomineralization rates of DOC measured in Lake Pääjärvi (markers) and the exponential attenuation of photomineralization (lines) after fitting the data to equation (4). The fitted photomineralization was 17.2 e−16 z and 16.7 e−13 z on 25 July and 2 August 2007, respectively.

Table 1. Measured and Modeled Rates (mmol C m−2 d−1) for the Photomineralization of DOC Over the Entire Water Column in Lake Pääjärvi at Different Temperatures, the Mean Temperature Over the in Situ Incubation (T), and the Activation Energy (Ea) Calculated for the Photomineralization of DOC
 25 Jul2 Aug
  • a

    Parameters c and d of the modeled ϕλ were 1.2374 and 0.024 nm−1, respectively, while ϕ330 was 450 × 10−6 mol C (mol photons)−1.

  • b

    Temperature correction calculated according to the mean Ea of 20.7 kJ mol−1.

  • c

    Mean temperature at a depth of 0.4 m measured with a temperature logger (HoboWare) at 60 min intervals during the in situ experiments (n = 24).

Measured rate (in situ)1.0791.280
Modeled ratea (+5°C)0.6710.929
Modeled rateb (in situ)1.0231.340
T in situc (°C)18.816.9
Eab (kJ mol−1)23.218.1

2.7. Modeling of Photochemical Rates in the Baltic Sea

[23] To calculate the photochemical rates at the surface of the Baltic Sea, we applied the measured aCDOM,λ, the measured ϕλ, and estimated Qs,z,λ to equation (1). The daily vector photon flux density spectrum at a depth of 0 m (Qv,0; mol photons m−2 d−1 nm−1) was estimated from an earlier radiative transfer model [Kuivikko et al., 2007] at latitude 60°N for different seasons (summer being June–August; autumn being September–November; winter being December–February; spring being March–May). The estimated Qv,0 was approximated to equal the Qs,z of equation (1) at the very surface of the Baltic Sea. The same Qv,0,λ was thought to equal the Qλ of equation (2). The ratio aCDOM atot−1 of equation (2) was assumed to be 1.

[24] To extend the salinity-dependent rates of the photoproduction of NH4+ and DIC over the entire Baltic Sea, we first calculated ϕNH4,λ or ϕDIC,λ for each salinity unit found in the surface water of the Baltic Sea. Then we estimated the geographical areas (km2) covered by each salinity unit with ArcMap software (ESRI) from the isohalines reported for the surface water in June (Figure 1) [Voipio, 1981] and summed up the extrapolated rates over the Baltic Sea. Because ϕbact-C,λ was independent of salinity, the mean seasonal rates per m2 of the studied stations were extended over the entire area of the Baltic Sea.

2.8. Temperature Correction

[25] We corrected the modeled photodecomposition rates (equations (1) and (2)) with the mean seasonal temperatures of the surface water. In the southern and central Baltic Sea, the mean temperatures are 16°C, 9°C, 3°C, and 7°C for summer, autumn, winter, and spring, respectively [Leppäranta and Myrberg, 2009]. Therefore, the correction factors with respect to 5°C were 1.40, 1.14, 0.94, and 1.07 for the summer, autumn, winter, and spring rates, respectively, according to an Ea of 20.7 kJ mol−1 applied to the Arrhenius equation (equation (5)). For the Gulf of Finland and the Gulf of Bothnia the mean summer, autumn, winter, and spring temperatures are 16°C, 9°C, 0°C, and 4°C at the surface water (Alg@line monitoring data), resulting in correction factors of 1.40, 1.14, 0.85, and 0.97, respectively. We assumed the same activation energy (Ea of 20.7 kJ mol−1) for the three photoreactions examined.

3. Results

3.1. Concentrations of DOC and DON

[26] The concentrations of DOC and the absorption coefficients of CDOM at 300 nm (aCDOM,300) decreased toward higher salinity, but the concentrations of DON were similar along the salinity gradient (Table 2). The C:N ratio of DOM was therefore highest in Neva Bay and decreased toward the more saline waters (Table 2). The aCDOM,300, the spectral slope coefficients (S), and the concentrations of DOC and DON (Table 2) were not statistically different between the seasons (repeated ANOVA; n = 3; p > 0.05) at salinities >5.

Table 2. Salinities, CDOM Absorption Coefficients (aCDOM) at 300 nm, Spectral Slope (S) at 300–550 nm, Concentrations of DOC and DON as Averages of Three Seasons, and the DOC:DON Ratio at the Sampling Stationsa
StationSalinityaCDOM,300 m−1S μm−1DOC mmol C m−3DON mmol N m−3DOC: DON
  • a

    Mean ± SD of the replicates, n = 1 for salinity, and n = 3 for other measurements.

Arkona Sea7.94.9 ± 0.116.6 ± 0.8334 ± 2214.7 ± 0.423 ± 2
Gotland Basin6.55.7 ± 0.417.2 ± 1.3371 ± 2214.5 ± 0.526 ± 2
Gulf of Finland5.88.0 ± 1.317.4 ± 1.3392 ± 3715.5 ± 0.825 ± 4
Helsinki5.19.9 ± 1.117.4 ± 0.9424 ± 3116.1 ± 0.626 ± 3
Neva Bay1.024.8 ± 3.416.3 ± 0.5623 ± 4615.1 ± 0.841 ± 5

3.2. Photochemical Production of NH4+, DIC and Labile DOM Supporting Bacterial Biomass

[27] The irradiations increased the concentrations of DIC and NH4+ (Table 3). The differences in concentrations between the irradiated and the dark waters were significantly higher than those between the dark and the initial waters (ANOVA; n = 5–7; p < 0.05), indicating that photochemistry was responsible for the mineralization of DOC and DON. The photoproduction of DIC decreased from Neva Bay toward the Arkona Sea, similar to aCDOM,300, but such a trend was lacking from the photoammonification (Table 3).

Table 3. The Photoproduction of NH4+, DIC, and Labile DOM Supporting Bacterial Biomassa
TimeStationNH4+ mmol N m−3 (n = 5)DIC mmol C m−3 (n = 7)Bacterial Biomass mmol C m−3 (n = 20)
  • a

    Mean ± SD of the replicates; n.d., not determined. Irradiation times varied between months, so the magnitudes of photoproduction are not directly comparable.

  • b

    Bacterial incubation samples were contaminated by heterotrophic nanoflagellates.

Jul 2006Arkona Sea0.25 ± 0.14n.d.b
 Gotland Basin0.34 ± 0.10n.d.4.0 ± 1.1
 Gulf of Finland0.37 ± 0.10n.d.2.6 ± 0.6
 Helsinki0.49 ± 0.11n.d.2.9 ± 0.7
 Neva Bay0.49 ± 0.05n.d.9.4 ± 2.5
 Irradiation time (h)66.745.9
Sep 2006Arkona Sea0.32 ± 0.158.1 ± 2.31.4 ± 0.3
 Gotland Basin0.57 ± 0.1912.3 ± 3.71.6 ± 0.4
 Gulf of Finland0.23 ± 0.1520.4 ± 2.71.3 ± 0.4
 Helsinki0.30 ± 0.0723.5 ± 1.11.7 ± 0.4
 Neva Bay0.50 ± 0.2265.1 ± 1.2b
 Irradiation time (h)69.168.346.1
Mar 2007Arkona Sea0.16 ± 0.0210.6 ± 2.52.0 ± 0.4
 Gotland Basin0.27 ± 0.0810.8 ± 4.02.2 ± 0.4
 Gulf of Finland0.29 ± 0.0413.0 ± 8.71.4 ± 0.3
 Helsinki0.33 ± 0.0617.2 ± 2.32.2 ± 0.5
 Neva Bay0.50 ± 0.0769.6 ± 3.310.0 ± 1.9
 Irradiation time (h)70.868.254.8

[28] Bacteria grew faster and achieved higher density in the irradiated waters than in the dark controls, as exemplified for the Gotland Basin in September (Figure 3). Additionally, bacterial cells were larger in the irradiated than in the dark control water (data not shown). The bacterial biomass was 1.6 mmol C m−3 higher in the irradiated than in the dark water at the highest cell density for the sample shown in Figure 3 (Table 3). For all samples, irradiations increased the bacterial biomasses by 1.3–10 mmol C m−3, or by 77%–765% compared to the biomasses in the dark waters (Table 3). Altogether, the simulated solar radiation mineralized DOC and DON as well as produced substrates that increased bacterial biomass in all stations and seasons (Table 3).

Figure 3.

Bacterial cell densities in the irradiated (L) and the dark samples (D) from Gotland Basin in September 2006 during incubation in the dark at 22°C–24°C. The photoproduction of labile DOM supporting bacterial biomass was calculated at the maximum cell density of the irradiated sample (long arrow). Error bars describe the standard error of the cell densities.

3.3. Apparent Quantum Yield Spectra for the Photoreactions

[29] The amount of photoproduced NH4+, DIC, and labile DOM supporting bacterial carbon biomass (Table 3) was normalized with the amount of absorbed photons (equation (S1)) for the determination of apparent quantum yield ϕλ (equation (3)). The ϕλs for the photoproduction of NH4+ and DIC showed no significant difference between the seasons (ANOVA; n = 5–7; p > 0.05). For each of these photoreactions the data from seasonal irradiation experiments were pooled, and mean ϕλs were iterated from pooled data (Table 4; equation (3)). The mean ϕNH4,λ and ϕDIC,λ depended on salinity (Figure 4 and Table 4; regression analysis for ϕNH4 F1,3 = 7.49, p = 0.072 and for ϕDIC F1,3 = 62, p = 0.004). For example, at a wavelength of 330 nm where photoreaction rates were the highest, ϕDIC,330 decreased with increasing salinity, whereas ϕNH4,330 behaved inversely (Figure 4 and Table 4). Therefore, the ϕDIC,330: ϕNH4,330 ratios were 19–58 and 112 in waters with salinities >5 and <5, respectively (Table 4).

Table 4. The Parameters c (Dimensionless) and d (nm−1) of the Apparent Quantum Yield Spectra ϕλ = c e (Equation (3)) for the Photoproduction of NH4+ and DIC, as Well as ϕNH4 and ϕDIC at 330 nm (×10−6 mol product (mol absorbed photons)−1) as Pooled Averages Across Three (NH4+) or Two (DIC) Seasons
StationPhotoproduction of NH4+Photoproduction of DICϕDIC:ϕNH4 at 330
cdϕ330CV%acdϕ330CV%a
  • a

    CV% (n = 3 for NH4+; n = 2 for DIC) describes the coefficient of variation for ϕλ based on the combined variability of the factors (prz, Qs,z,λ, aCDOM,λ; equation (1)) used for the determination of ϕλ.

Arkona Sea0.85030.03498.5181.12870.02681631719
Gotland Basin0.89220.034211.2121.15090.02582312521
Gulf of Finland0.80050.03615.491.17860.02542703350
Helsinki0.74840.03644.561.17210.0255260258
Neva Bay0.72040.03713.5191.19160.02433926112
Figure 4.

The mean apparent quantum yields at a wavelength of 330 nm (ϕ330) across three seasons for photoammonification (NH4+), the photoproduction of DIC (DIC), and labile DOM supporting bacterial biomass (bact-C) in the salinity gradient under study. The coefficient of determination (R2) describes the linear fit between ϕ330s and salinity.

[30] The ϕλs for the photoproduction of labile DOM supporting bacterial biomass (ϕbact-C) differed significantly between the seasons (ANOVA; n = 3; p < 0.05; Table 5). Unlike ϕDIC,330 and ϕNH4,330, the mean ϕbact-C,330 over the seasons did not depend on salinity (Figure 4 and Table 5; regression analysis F1,3 = 0.11, p = 0.764). The mean ϕbact-C,330 corresponded to 14%–32% of ϕDIC,330 (Figure 4).

Table 5. The Parameters c (Dimensionless) and d (nm−1) of the Apparent Quantum Yield Spectra ϕbact-C,λ = c e (Equation (3)) for Photoproduction of Labile DOM Supporting Bacterial Biomass, ϕbact-C at 330 nm ( × 10−6 mol C (mol absorbed photons)−1), and Coefficient of Variation (CV%) for the Individual Determinations in March, July, and September
TimeStationcdϕ330CV%a
  • a

    See Table 4. N.d., not determined.

Mar 2007Arkona Sea0.99970.0304444
 Gotland Basin1.03610.02936611
 Gulf of Finland1.00040.0313336
 Helsinki1.00800.03093811
 Neva Bay1.03530.02985611
Jul 2006Arkona Sean.d.n.d.n.d.n.d.
 Gotland Basin1.10550.027911110
 Gulf of Finland1.05350.0298579
 Helsinki1.03370.02985513
 Neva Bay1.05920.02936722
Sep 2006Arkona Sea1.01090.03005112
 Gotland Basin0.99770.03044416
 Gulf of Finland0.90900.03212315
 Helsinki0.94500.03182612
 Neva Bayn.d.n.d.n.d.n.d.

3.4. Photoreaction Rates During a Summer Day

[31] To calculate the rates of photochemical reactions at the sampling stations, the apparent quantum yield spectra for the photoreactions (ϕλ; Tables 4 and 5), the measured absorption coefficient spectra of CDOM (aCDOM), and the mean seasonal solar irradiation spectra (e.g., for a summer day, shown in Figure 5a) were applied to equations (1) and (2). Although the solar irradiance increased toward the visible part of the spectrum (Figure 5a), the aCDOM (Figure 5b) and ϕλs (Figure 5c) increased exponentially toward the UV range of the spectrum, which was largely responsible for the photoreactions (Figures 5d–5f). All three photoreaction rates peaked at a wavelength of 330 nm both at the surface and over the water column (Figures 5d–5f). In the Gulf of Finland, for example, after the integration across the spectra (Figure 5e), the photoproduction rates of DIC were 1.6 mmol C m−3 d−1 at the surface and 0.72 mmol C m−2 d−1 over the water column in summer (Figures 6b and 6e).

Figure 5.

The photoproduction rates of NH4+, DIC, and labile DOM supporting bacterial carbon modeled at the “Gulf of Finland” station in summer and the parameters needed for the models (equations (1) and (2)). (a) A mean daily vector photon flux density spectrum (Qv,d,λ) in summer at latitude 60°N. (b) The measured absorption coefficient spectrum of CDOM (aCDOM,λ). (c) The apparent quantum yield spectrum for the photoproduction of NH4+, DIC, and bacterial carbon (ϕNH4,λ, ϕDIC,λ, ϕbact-C,λ from Tables 4 and 5). The modeled photoproduction rates of (d) NH4+, (e) DIC, and (f) bacterial carbon in situ at the surface (μmol NH4+/C m−3 d−1; black line) and over the entire water column (μmol NH4+/C m−2 d−1; gray line).

Figure 6.

Summer rates of photochemical reactions (a–c) at the surface layer and (d–f) over the entire water column of the stations. The error bars describe the standard deviation of the method calculated with combined coefficients of variation from the determination of aCDOM,λ and the photoreactions according to Wilkinson [1961].

[32] When examined over the entire water column, the photomineralization of DOC and DON followed the salinity dependence of ϕNH4,330 and ϕDIC,330 (Figure 4 and Figures 6d–6e). The photoproduction rates of DIC were highest in Neva Bay and decreased toward the Arkona Sea (Figure 6e). The photoproduction rates of NH4+ were lowest in Neva Bay and generally increased toward the Arkona Sea (Figure 6d). The photoproduction rate of DIC exceeded severalfold the other rates examined (Figure 6). For example, during summer in the Gulf of Finland, the photoproduction rate of DIC was 0.72 mmol C m−2 d−1, whereas the photoproduction of labile DOM supporting bacterial biomass was 0.13 mmol C m−2 d−1 (i.e., 18% of the photoproduction of DIC; Figures 6e–6f), and the photoammonification rate was 0.0098 mmol NH4+ m−2 d−1 (Figure 6d).

3.5. Seasonal and Annual Photoreaction Rates

[33] The annual photoreaction rates were calculated as a sum of seasonal rates based on seasonal solar irradiances and mean temperatures (Figure 7). The photochemical decomposition of DOM was highest during summer with the highest solar irradiance and water temperatures. The summer rates of photoproduction of NH4+ and DIC contributed 55%–57% to the annual rates (Figures 7a–7b). Because ϕbact-C,λ was greatest in summer (Table 5), the photoproduced labile DOM during summer contributed 56%–73% to the annual photoproduction of labile DOM supporting bacterial biomass (Figure 7c). In winter, the photochemical reactions were limited by low light intensity and slowed further due to low temperature.

Figure 7.

Sum of seasonal photoproduction rates of (a) NH4+, (b) DIC, and (c) bact-C over the entire water column (error bars as in Figure 6).

[34] The annual photoreaction rates per square meter (Figure 7) were extrapolated to the annual rates across the entire Baltic Sea taking into account the salinity dependence of the photoreaction rates. The annual rate of photoammonification was 0.043 Tg N yr−1 and its range from 0.038 to 0.049 Tg N yr−1 when calculated accounting the median CV of 12% for the determination of ϕNH4 (Table 6). The photochemical reactions mineralized 1.90 Tg DOC yr−1 to DIC (range 1.57 to 2.24 Tg DOC yr−1 based on median CV of 17% for ϕDIC) and produced labile substrates, which supported 0.38 Tg C yr−1 of bacterial biomass (range 0.34 to 0.43 Tg C yr−1 based on median CV of 11% for ϕbact-C; Table 6).

Table 6. Annual Rates for Photoreactions and Other Selected Nitrogen and Carbon Fluxes With Estimated Ranges in the Baltic Sea (Surface Area of 377,005 km2)
1012 g C (Baltic Sea)−1 yr−1ProcessReference
  • a

    The range is calculated with a median CV% of the sampling stations determined for each photoreaction (see Tables 4 and 5).

  • b

    Annual river loading of total N subtracted by the annual river loading of NO3N.

  • c

    Annual rates in the Baltic Proper (area of 244,000 km2) extended over the entire Baltic Sea.

  • d

    Calculated from the photoproduced bacterial biomass based on the phototransformed DOM with a bacterial growth efficiency of 0.20–0.25 determined for bacterial growth on photoproducts [Pullin et al., 2004] and for bacterioplankton in the Baltic Sea [Hagström et al., 2001], respectively.

  • e

    Annual mean in the northern Baltic Sea extended over the entire Baltic Sea.

0.04–0.05aDON photochemistry → NH4+This study
0.21–0.29Atmospheric deposition of inorganic NBartnicki et al. [2007]; Thomas et al. [2010]
0.55–0.83River loading of total NStålnacke et al. [1999]; Thomas et al. [2010]
0.28–0.41River loading of DONbStålnacke et al. [1999]; Thomas et al. [2010]
0.09–0.19N2 fixation by cyanobacteriacDegerholm et al. [2008]
0.78–1.03Sedimentation and denitrificationSavchuk [2005]; Thomas et al. [2010]
1012 g N (Baltic Sea)−1 yr−1ProcessReference
1.57–2.24aDOC photochemistry → DICThis study
0.34–0.43aDOC photochem. → bacterial biomassThis study
0.79–1.28aDOC photochem. → bacterial respirationdThis study
0.14–0.36eAtmospheric deposition of TOCAnttila et al. [1995]; Thomas et al. [2010]
2.11–4.90eRiver loading of TOCAlgesten et al. [2006]; Thomas et al. [2010]
3.59e–18.1Sedimentation of TOCElmgren [1984]; Algesten et al. [2006]
9.34–12.6Bacterial productioneElmgren [1984]; Sandberg et al. [2004]
37.2–49.8Phytoplankton productionElmgren [1984]; Thomas et al. [2010]
1.79–4.69Mesozooplankton productioneElmgren [1984]; Sandberg et al. [2004]
0.08–0.12Fish catchElmgren [1984]; Thomas et al. [2010]

4. Discussion

4.1. Apparent Quantum Yields

[35] To our knowledge, this is the first study to determine apparent quantum yields (ϕλ) for the photoproduction of DIC, NH4+, and labile DOM supporting bacterial biomass from the same water samples. This and many previous studies describe ϕλs with two parameters, resulting in spectra which decrease exponentially toward the longer wavelengths [Johannessen and Miller, 2001; White et al., 2010]. Some studies use three parameters for ϕλ, resulting in a slightly different shape for the spectrum of ϕ [Bélanger et al., 2006; Yang et al., 2011]. For the comparison of our ϕλs with previous published ϕλs modeled with different approaches [Bélanger et al., 2006; Yang et al., 2011], we selected the ϕ at a wavelength of 330 nm, where the photoreaction rates are highest [Bélanger et al., 2006; White et al., 2010; this study].

[36] Previously published ϕDIC,330 values range from 136 × 10−6 at a salinity of 8.2 [Bélanger et al., 2006] to ∼2000 × 10−6 in river water [Gao and Zepp, 1998], thus covering the range from 163 × 10−6 (Arkona Sea) to 450 × 10−6 mol C mol photons−1 (Lake Pääjärvi) as determined in this study. Generally, previous ϕDIC,330 values determined for terrestrial DOM in lakes and rivers are 2–12 fold higher than those of our Baltic Sea samples [Gao and Zepp, 1998; Vähätalo and Wetzel, 2004; White et al., 2010]. In coastal waters, such as the Delaware Estuary and the Mid-Atlantic Bight, ϕDIC,330 values are similar to our Baltic Sea values [Johannessen and Miller, 2001; White et al., 2010], but in the Mackenzie Shelf, ϕDIC,330 values measured at 0°C are only 18%–33% of our values [Bélanger et al., 2006]. In conclusion, our ϕDIC,λ for Baltic Seawater support previous findings that the photochemical reactivity of DOC is generally higher in fresh than in coastal waters.

[37] The ϕNH4,330 values of the present study are lower than previous ϕ330 values for the photoproduction of NH4+ and labile N in the Baltic Sea [Vähätalo and Zepp, 2005; Stedmon et al., 2007; Vähätalo and Järvinen, 2007]. At salinities >5, previous ϕNH4,330 values range from 11.4 × 10−6 to 142.4 × 10−6 mol NH4+ mol photons−1 [Vähätalo and Zepp, 2005; Stedmon et al., 2007], and ϕlabile-N,330 values from 7.8 × 10−6 to 16.6 × 10−6 mol N mol photons−1 [Vähätalo and Järvinen, 2007; Vähätalo et al., 2011]. At salinities <5, the ϕNH4,330 values determined here are only 6%–13% of previously published ϕNH4,330 values [Vähätalo and Zepp, 2005; Stedmon et al., 2007]. The low temperature (+5°C) used in this study prevented microbial interference during the irradiation, but obviously slowed photoammonification and possibly contributed to the low ϕNH4,λ values found in this study. Additionally, the high doses of absorbed photons used in this study may have contributed to our low ϕNH4,λ. If only a small portion of DON can be converted to NH4+, ϕNH4,λ will be highest at the lowest doses of irradiation, but will decrease with the depletion of photoreactive DON and increasing doses of absorbed photons [Bushaw et al., 1996; Kitidis et al., 2008; Vähätalo and Zepp, 2005]. In the case of ϕDIC,λ, photoreactions can generate new carboxylic groups, which can later be converted to DIC through photodecarboxylation [Xie et al., 2004]. Such regeneration of photoreactive moieties may be lacking from photoammonification, which can lead to a decrease in ϕNH4,λ with longer irradiation times.

[38] To the best of our knowledge, besides a single earlier determination from the Baltic Sea [Vähätalo et al., 2011], no apparent quantum yields for the photoproduction of labile DOM supporting bacterial biomass have been published previously. In the present study, the ϕbact-C,330 values range from 23 × 10−6 to 111 × 10−6 mol C mol photons−1, which agree with the value of 91 × 10−6 mol C mol photons−1 determined earlier for the northern Baltic Sea [Vähätalo et al., 2011]. The ϕbact-C,330 values correspond to only 2%–10% of previously published ϕ330 values for bacterial respiration based on photoproduced DOM determined in coastal Georgia [Miller et al., 2002]. Different bacterial growth efficiencies (BGE), however, may explain how photoproduced DOM contributes to the stimulation of bacterial biomass and respiration in different environments. The variability in BGE of bacteria grown with photoproduced DOM may be high (e.g., ranging from 0.02 to 0.41 among different photoproduced substrates) [Bertilsson and Tranvik, 1998; Vähätalo et al., 2003; Pullin et al., 2004]. Our ϕbact-C,λ values allow one to calculate directly the photoproduction of labile DOM supporting biomass and its contribution to heterotrophic production without determining BGE.

[39] The stations selected for this study represent a transition from allochthonous DOM at Neva Bay toward autochthonously produced marine DOM in the Arkona Sea away from direct riverine sources. The changes in the photochemical reactivities (i.e., ϕλ) can therefore be examined in relation to the shift in the source of DOM. In this study, the photoreactivity of DON was greatest in higher salinities, as reported previously for the Baltic Sea [Stedmon et al., 2007] and for the Orinoco River plume [Morell and Corredor, 2001]. In contrast, the photoreactivity of DOC was greatest in low salinities, which agrees with previous results from the Mackenzie Shelf and Delaware Estuary [Bélanger et al., 2006; White et al., 2010], but disagrees with results from the Gulf Stream, Mid Atlantic Bight, and Halifax Harbour [Johannessen and Miller, 2001]. Concluding from this and most previous studies, terrestrial DOC seems to be more photoreactive than marine DOC, but in terms of photoammonification, marine DON is more reactive than terrestrial DON.

[40] When our ϕDIC:ϕNH4 ratios are compared to the C:N ratio of DOM, the values are 0.8- to 2.2-fold in the open Baltic Sea, but 2.7 times higher in Neva Bay. This comparison shows that solar radiation photomineralizes more DOC than DON, especially in low-saline waters. The photochemical transformation of DOM differs from the microbial consumption of DOM [Lønborg et al., 2010]. Microbes prefer DON to DOC, which tends to increase the C:N ratio of DOM. Because the C:N ratio of DOM typically decreases from terrestrial sources toward the marine environment [Bronk, 2002], the photochemical transformation is likely a major contributor to this decrease in coastal waters.

4.2. Seasonality

[41] According to our results, seasonality has no impact on the concentration of biologically recalcitrant DOC or DON nor the photochemical reactivity (ϕDIC,λ and ϕNH4,λ) of such DOM in the open Baltic Sea. The lack of seasonality in our ϕDIC,λ and ϕNH4,λ was somewhat unexpected, because the photochemical bleaching of chromophoric dissolved organic matter (CDOM) can decrease the photochemical reactivity of DOM during the most intensive radiation in summer, at least in a small humic lake [Vähätalo et al., 2003]. In the Arkona Sea, we estimate that solar radiation mineralizes 260 mmol C m−2 and 10.2 mmol N m−2, and decreases the concentration of DOC by 5.2%–7.9% and that of DON by 4.6%–6.9% in the 10–15 m deep mixing layer between March and September. A previous study in a freshwater reservoir found that the apparent quantum yield for photobleaching decreases 17% when CDOM absorbance photobleaches by 50% [Vähätalo and Wetzel, 2004]. In the Arkona Sea, the calculated photochemical decrease in concentrations of DOC and DON was less than 7.9% (see above). Assuming that the photoreactivity of DOM in the Arkona Sea decreases as reported [Vähätalo and Wetzel, 2004], the photochemical reactivity of DOM decreases by less than 2.7% between March and September. Such a small seasonal decrease in photochemical reactivity is below the accuracy of the method used in this study, and therefore explains the lack of seasonality in ϕDIC,λ and ϕNH4,λ.

[42] In contrast to ϕDIC,λ and ϕNH4,λ, we found that seasonality contributes to ϕbact-C. As the abiotic photoreactions (the photomineralization of DOC and DON) lacked seasonal variability, the seasonal variation of ϕbact-C,λ likely stems from seasonal differences in bacterial communities. Bacterial communities go through an annual succession where different taxonomic and functional groups are present at different times of the year [Salcher et al., 2008]. For example, the genus Polynucleobacter is most abundant during summer, with the highest solar irradiance [Wu and Hahn, 2006]. Polynucleobacter has a high affinity for small organic acids and photoproduced labile DOM substrates [Watanabe et al., 2009]. Therefore, the highest ϕbact-C,λ during summer may stem from the seasonal occurrence of bacterial strains with a high affinity for photoproduced DOM.

4.3. Evaluation of Errors

[43] We made many assumptions and simplifications when calculating the rates of phototransformations. First, the variability in the determination of ϕλs raises uncertainty, represented by CV% in Tables 4 and 5 and shown as error bars in Figures 6 and 7.

[44] Second, the assumption that CDOM absorbs all radiation (i.e., aCDOM atot−1 = 1; equation (2)) in the Baltic Sea is not entirely correct, because particles also absorb solar radiation. In the southern Baltic Sea, the aCDOM atot−1 ratio (equation (2)) at 380 nm is 0.8 [Babin et al., 2003]. However, the ratio is higher at the shorter photochemically important wavelengths than at 380 nm [Uusikivi et al., 2010]. The absorption by CDOM increases from the southern toward the northern Baltic Sea (Table 2), and therefore the aCDOM atot−1 ratio at photochemically important wavelengths is likely larger than 0.8 reported for the southern Baltic Sea at 380 nm. As our photoreaction rates concerning CDOM did not account for the absorption by particles, the rates are slightly overestimated. The magnitude of overestimation is perhaps likely as small as a few percent but less than 20% at maximum.

[45] The Ea determined in this study estimates the temperature dependence for the photoproduction of DIC from natural surface waters for the first time. As our estimate is based on the comparison of two temperatures through modeling, it should be considered only as a crude estimate. Our Ea is, however, similar to the photoproduction of DIC from a single substance, pyruvic acid, and in the middle of range (from 10 to 30 kJ mol−1) reported for Ea of aqueous photoreactions [Turro, 1991; Schwarzenbach et al., 2003; Guzmán et al., 2007]. We used the same Ea determined for the photoproduction of DIC to correct also the photoammonification and the photoproduction of labile DOM supporting bacterial biomass, as this information was the best available. Consequently, our rate estimates for the latter reactions are less accurate than for the photoproduction of DIC. Zhang et al. [2006] found that Ea for the photoproduction of CO is 78% higher in fresh waters than in marine waters. Although such salinity dependence is unknown in the photoproduction of DIC, our Ea determined in lake water may overestimate photoreactions in the Baltic Sea. Overall, the temperature dependence of photochemical reactions is low compared to thermal reactions [Turro, 1991; Schwarzenbach et al., 2003]. For example, using an Ea of 10 kJ mol−1 changed the temperature correction factors closer to 1 in this study, thus decreasing the phototransformation rates by 3%–16% when the in situ temperature was >5°C, and increasing the rates by 2%–8% at the in situ temperatures <5°C. Even though the in situ temperatures were carefully taken into account in the validation calculations (correction factors 0.85–1.40 in this study), the amount of irradiation is, in general, a more important factor affecting the photochemical reaction rates than is temperature.

4.4. Environmental Relevance of DOM Photochemistry

[46] Unlike in most coastal waters, we can calculate the phototransformation rates within the distinct boundaries of the Baltic Sea and compare them to the other C and N fluxes determined earlier for this coastal sea (Table 6). The annual C and N fluxes shown in Table 6 are intended to place the phototransformation of DOM into perspective with other biogeochemical processes. Many simplifications have been made for extrapolating the fluxes found in the literature to cover the entire Baltic Sea and the annual cycle.

[47] The photoproduced labile DOM supporting bacterial biomass links photochemically transformed DOC to higher trophic levels [Vähätalo et al., 2011]. The annual photostimulation of bacterial carbon biomass contributes 3%–5% to total annual bacterial production (Table 6) [Elmgren, 1984; Sandberg et al., 2004]. The magnitude of bacterial biomass based on photoproduced labile DOM corresponds to 0.7%–1.1% of primary production, 7%–24% of mesozooplankton production, or exceeds the fish catch in the Baltic Sea three- to sixfold (Table 6) [Elmgren, 1984; Sandberg et al., 2004; Thomas et al., 2010]. The magnitude of photoammonification corresponds to 13%–23% of the atmospheric deposition of inorganic N, 5%–9% of the river loading of total N, 9%–18% of the river loading of DON, 4%–6% of the denitrification and sedimentation of total N [Stålnacke et al., 1999; Savchuk, 2005; Bartnicki et al., 2007; Thomas et al., 2010], or 20%–56% of N2 fixation by cyanobacteria (Table 6) [Degerholm et al., 2008].

[48] The magnitude of the direct photomineralization of DOC corresponds to 32%–106% of the river loading of total organic carbon (TOC) or 9%–62% of the sedimentation, and exceeds the atmospheric deposition of TOC to the Baltic Sea more than fourfold (Table 6) [Elmgren, 1984; Anttila et al., 1995; Algesten et al., 2006; Thomas et al., 2010]. The photochemical transformations not only directly mineralize DOC to DIC, but also produce labile DOM for bacterioplankton. In our study, the photoproduction of labile DOM supporting bacterial biomass corresponds to 20% of the direct photomineralization of DOC. However, the total photoproduction of labile organic substrates is higher than the bacterial biomass measured in this study, because part of the photoproduced labile carbon is lost through bacterial respiration. When bacterial growth efficiencies (BGE) of 0.20–0.25, as determined earlier for the Baltic Sea [Hagström et al., 2001] and for phototransformed DOM in lake and river waters [Vähätalo et al., 2003; Pullin et al., 2004], is taken into account, the total photoproduction of bioavailable substrates accounts for 72%–76% of the direct photomineralization of DOC. The total photochemical transformation of DOM through direct and indirect mineralization by bacteria amounts to 2.71–3.94 Tg C yr−1, which approximately equals the annual riverine input of allochthonous TOC to the Baltic Sea (2.11–4.90 Tg C yr−1) (Table 6) [Algesten et al., 2006; Thomas et al., 2010].

[49] The photochemical reactions can mineralize only the photoreactive portion of DOC. Previous studies indicate that photochemical reactions (directly and assisted by microbes) can mineralize 80%–97% of wetland-derived DOC and 36%–41% of lake DOC [Obernosterer and Benner, 2004; Vähätalo and Wetzel, 2004, 2008]. Assuming that half of allochthonous TOC is photoreactive, total photochemical mineralization (2.71–3.94 Tg C yr−1) in the Baltic Sea exceeds the annual input of photoreactive allochthonous DOC (<2.45 Tg C yr−1) to the Baltic Sea. Our estimate is similar to those done earlier for the global coastal ocean, where the carbon losses through photoreactions exceed riverine inputs [Miller et al., 2002; Wang et al., 2009]. Because the photochemical losses cannot be larger than the sources of photochemically reactive DOM, photochemical reactions must transform also autochthonous photoreactive DOM produced in the coastal waters.

[50] Our finding that solar radiation can completely mineralize photoreactive terrestrial DOC in the Baltic Sea contrasts with the corresponding estimate of Bélanger et al. [2006] in the coastal Arctic Ocean. Bélanger et al. [2006] estimated that only a few percent of terrestrial DOC is photomineralized in a coastal section of the Arctic Ocean with an area corresponding to that of the Baltic Sea. Several factors explain the difference in phototransformation between the Arctic Ocean and the Baltic Sea. During a typical year, only 25% of the section of Arctic Ocean under study is ice-free, thereby substantially limiting the phototransformations in the water column. Annual doses of solar radiation and temperatures are lower in the Arctic Ocean than in the Baltic Sea. Unlike in our study, Bélanger et al. [2006] did not account for the biological mineralization of phototransformed DOC. Further, a deep water formation and an export below the photolytic surface layer are relevant sinks for the terrestrial DOC of Arctic rivers. In the Baltic Sea and other coastal seas at latitudes below 60°N, terrestrial riverine DOC remains in the mixed surface layer and becomes exposed to extensive doses of photolytic solar radiation. Altogether, our study provides quantitative support for the previous suggestion of Miller and Zepp [1995] that photochemical transformation is a major sink for terrestrial riverine DOC in the mid- and low-latitude coastal ocean.

Acknowledgments

[51] We thank the Alg@line project for supplying the water samples from stations from Helsinki to the Arkona Sea; and Pauli Haimi, Jouni Lehtoranta, and Jyrki Vuorinen for providing the water samples from Neva Bay. We also thank the laboratory staff at the Tvärminne Zoological Station and Marine Research Center for the DOC/DON and nutrient analyses, Martí Galí Tàpias for his assistance in measuring CDOM absorbance, and Veijo Kinnunen for drawing the Baltic Sea map. Funding was provided by the Academy of Finland.

Ancillary