Biologically labile photoproducts from riverine non-labile dissolved organic carbon in the coastal waters

V. Kasurinen, H. Aarnos, and A. Vähätalo University of Helsinki, Department of Forest Sciences, P.O. Box 27, Helsinki, Finland Norwegian University of Science and Technology, 7491 Trondheim, Norway University of Helsinki, Department of Environmental Sciences, P.O. Box 65, 00014 Helsinki, Finland University of Jyväskylä, Department of Bio and Environmental Sciences, P.O. Box 35, 40014, Jyväskylä, Finland


Introduction
Rivers transport 246 Mt C yr −1 of dissolved organic carbon to the ocean Cai (2011). A part (19 ± 16 %) (Søndergaard and Middelboe, 1995) of this terrestrial DOC is biologically labile and quickly consumed by heterotrophic bacteria in the coastal waters 25 (Lønborg et al., 2009; Lønborg and Álvarez Salgado, 2012). The remaining majority Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | (81 ± 16 %) of terrestrial DOC (tDOC) is biologically non-labile (Søndergaard and Middelboe, 1995). In the coastal waters, the primary sinks for non-labile DOC are slow microbial metabolism or photochemical transformations (Bauer et al., 2013). Photochemical reactions can transform non-labile tDOC to dissolved inorganic carbon (DIC) or biologically labile photoproducts (BLPs), the latter being rapidly assimilated by mi-5 crobes (Wetzel et al., 1995). The assimilation of non-labile tDOC (directly or after photochemical transformation) contributes to heterotrophy of coastal waters (Smith and Hollibaugh, 1993) and couples tDOC to marine food webs (Vähätalo and Järvinen, 2007;Vähätalo et al., 2011). The extent of this coupling depends also on the fraction of assimilated tDOC converted to bacterial biomass (i.e., bacterial growth efficiency, 10 BGE) (del Giorgio and Cole, 1998;del Giorgio and Davis, 2003).
By operational definition, BLPs are compounds produced by photochemical reactions and transformed into a more labile form than the original compounds in DOC. The amount of BLPs is typically quantified experimentally as bacterial respiration, bacterial production or associated loss of DOC (= bacterial carbon demand) in irradiated 15 DOC in excess of that found in non-irradiated dark control DOC (Wetzel et al., 1995;Smith and Benner, 2005). In order to translate the experimental findings to environmental processes, the amounts of BLPs have been related to the number of photons absorbed by chromophoric dissolved organic matter (CDOM) yielding to an apparent quantum yield (AQY) for production of BLPs (Miller et al., 2002;Vähätalo et al., 2011;20 Aarnos et al., 2012;Reader and Miller, 2014;Cory et al., 2014).
The production of BLPs is also linearly related to the photochemical loss of CDOM (photobleaching) (Kieber et al., 1990;Miller and Moran, 1997;Bertilsson and Tranvik, 2000;Obernosterer and Herndl, 2000;Brinkmann et al., 2003). Although, this quantitative relationship between the production of BLPs and photobleaching could be also 25 used to estimate the photoproduction of BLPs in the surface waters, it has to our knowledge never been used for such a purpose. Bacterial production on BLPs can support both heterotrophic food webs and primary production in the coastal waters (Vähätalo and Järvinen, 2007;Vähätalo et al., 2011). The bacterial production on BLPs depends 8201 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | in part on BGE on BLPs. Irradiation of DOC can decrease (15 studies) or increase BGE (5 studies) compared to non-irradiated DOC (reviewed by Abboudi et al. (2008)). Despite of these comparisons, between irradiated and non-irradiated DOC, the BGE on BLPs has not been determined and our study provides the first BGE estimate on BLPs. 5 This study estimates the production of BLPs from non-labile tDOC in the coastal waters and determines BGE on BLPs. We collected water samples from ten large rivers that produce together 28 % of tDOC flux to the ocean and aged the samples with indigenous microbes for the removal of labile DOC. The remaining non-labile DOC was mixed (1 : 1) with artificial seawater to simulate chemical matrix of coastal ocean. The 10 sterile filtered mixtures of non-labile DOC and seawater were irradiated with simulated solar radiation. The irradiated waters received indigenous bacteria and the amount of BLPs was measured separately as bacterial production (BP on BLPs) and respiration (BR on BLPs) in excess of that found in the non-irradiated dark control waters.The BP or BR on BLPs was related to the number of absorbed photons for the calculation of 15 AQY and also to the photobleaching (BLPs vs. photobleaching approach). The annual production of BP on BLPs and BR on BLPs (mol C m −2 yr −1 ) in the river plumes was calculated from AQYs and the annual solar radiation in the coastal ocean in the front of each river. We calculated the annual production of BLPs from the annual CDOM fluxes for each river (Mt C yr −1 per river) using the BLPs vs. photobleaching approach 20 and assuming that photobleaching of tDOM in the coastal ocean equals the annual tDOM flux. As the rivers examined contribute 28 % of tDOC flux to the ocean, we dared to extrapolate our finding to global scale simply by assuming that the rivers examined here are representative for the remaining 72 % of tDOC flux by other rivers.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 2 Material and methods

Material
The rivers selected for this study (Table 1) drain 25 % of the land area, contribute 28 % to the global tDOC flux, and 33 % to the freshwater discharge to the ocean (Carlson, 2002;Cauwet, 2002;Coynel et al., 2005;Milliman and Farnsworth, 2011). For 5 the collection of water samples, empty polyethylene containers (cleaned with detergent, rinsed with HCl and finally copiously rinsed with ion exchanged water; MQ) were shipped to the local collaborators at ten major rivers (see acknowledgments). The containers were filled in the center of the stream by direct immersion below the surface, except for the Mississippi and St. Lawrence Rivers where water about 3 m below the 10 surface was collected with Niskin bottles. The samples collected for this study has been used earlier for the determination of dissolved black carbon (Jaffé et al., 2013), dissolved iron (Xiao et al., 2013) and the photochemical isotopic fractionation of DOC (Lalonde et al., 2014). Artificial seawater was prepared according to Kester et al. (1967) by dissolving pro 15 analysis grade of inorganic salts (VWR International) to MQ water. Prior to use, glassware and quartzware were washed in a dishwasher, rinsed out with 7 % HCl and MQ water, and combusted for several hours (450 • C). 20 In order to remove labile DOC, the collected water samples were not filtered or preserved, but kept in darkness and sent to our laboratory in Helsinki, Finland by air cargo. The median transport time between the sampling and the arrival to Helsinki was 33 days (  dergaard and Middelboe, 1995;Lønborg and Álvarez Salgado, 2012;Asmala et al., 2014). Our salinity adjustment followed by filtration mimics a potential removal of iron via flocculation and related loss in photochemical reactivity and is expected to result a ma-10 trix characteristic for photochemical transformation of non-labile tDOC in the coastal waters 2.2.1. In order to assess BGEs and the production of BLPs from non-labile tDOC under conditions representing coastal waters, the 1 µm-filtered river water samples were mixed 1 : 1 with artificial seawater. In the mixtures, the ionic composition was similar to that of seawater but the salinity (16) was half of that found in the ocean. For 15 abiotic photochemical experiments, the 1 µm filtered river water samples mixed with artificial seawater were further aseptically filtered through 0.2 µm membrane filters (Palls Supor 200) into 170 mL glass or quartz flasks to be closed with ground-glass stoppers without headspace.

20
The quartz flasks were irradiated for 44 to 46 h with 765 W m −2 simulated solar radiation (Atlas Suntest CPS+ solar simulator) in a MQ-water bath regulated to 20 • C with a Lauda RE-112 thermostat like in our earlier study (Aarnos et al., 2012). The glass flasks wrapped into aluminum foil were kept in the same water bath (dark controls) or at 4 • C (initials). 25 8204 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Bioassays
For the bioassays, the irradiated and the dark control samples from each river received a 10 % (vol/vol) inoculum of bacterioplankton prepared from their corresponding river by passing the 1 µm-filtered river water through GF/F filters (nominal pore size of 0.7 µm, Whatman) for the removal of the grazers of bacterioplankton. The bioassay 5 samples received NH 4 Cl and KH 2 PO 4 to the final concentrations of 1.2 mg N L −1 and 0.13 mg P L −1 , respectively, and were divided into two aliquots for the determination of bacterial respiration (BR) or bacterial production (BP).

Bacterial respiration
BR was determined as a consumption of dissolved oxygen (O 2 ) measured with needle-10 type oxygen microsensor optodes (PreSens GmbH, Regenburg, NTH-PSt1-L5-TF-NS40) (Warkentin et al., 2007(Warkentin et al., , 2011 at 15 min intervals from aliquots closed in biological oxygen demand bottles incubated in darkness in a water bath at 20 • C (Lauda RE-112 thermostat). An optode was inserted into the sample via a hole drilled through the ground-glass stopper and sealed with parafilm. The drift of instrument defining 15 the detection limit for BR was measured in three blank experiments, where MQ water was incubated for 300 h under conditions identical to respiration measurements. During the blank experiments, the apparent decline of O 2 was 1.5 ± 0.5 µmol L −1 300 h −1 (mean ± sd, n = 3). The decline in the concentration of O 2 was converted into an increase in the concentration of CO 2 assuming 1 : 1 molar ratio between the consumed 20 O 2 and the produced CO 2 in BR. The selected respiratory quotient makes our study comparable to earlier studies, which have used values ranging from 0.82 to 1.2 (Søndergaard and Middelboe, 1995;del Giorgio et al., 1997;del Giorgio and Cole, 1998;Cory et al., 2014). The temporal trend of accumulated CO 2 was determined by a polynomial fitting to the measurements with the smooth spline function in R software (R

8205
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Bacterial production
To determine bacterial production (BP), 10 mL samples were preserved with formaldehyde (final concentration 5 %) daily from aliquots incubated in dark conditions at room temperature (mean 22 • C, range from 20 to 24 • C) with an air-head space. The bacterial cells from the preserved samples were filtered on 0.2 µm polycarbonate filters (GE 5 water & Process Technologies) and stained with acridine orange (Hobbie et al., 1977). Bacterial densities were estimated from cell counts from ≥ 20 fields with an epifluorescence microscope (Aristoplan). The cell volumes were determined by digital image analysis (Image Analysis & LabMicrobe software; Massana et al.1997). The volumes of cells were converted to carbon using a conversion factor of 0.12 pg C (µm 3 ) 0.7 and 10 the total carbon-biomass of bacterioplankton assemblage was calculated by multiplying the organic carbon content of cells with the density of cells (Norland, 1993).

Defining BLPs and BGEs
For the bacteria growing in the irradiated sample, the response to BLPs was considered to be largest in the end of logarithmic growth phase defined as the first day with 15 maximum biomass during the 12 day-long bioassay (Stepanauskas et al., 2002). On that day (varying from 5 to 12 days), the difference in the accumulated BP or BR between the dark control and the irradiated samples represented the BP or BR on BLPs. The bacterial growth efficiency (BGE) was calculated as BP/(BP + BR). 20 The absorbance by CDOM was measured with a spectrophotometer (Shimadzu UV 2550) in triplicates against MQ water reference using a 0.05 m quartz cuvette. The absorbance by CDOM was converted into an absorption coefficient (a CDOM,λ ) accounting for the path length of the cuvette and transforming the 10-base logarithm to a Naperianbased logarithm (ln(10)).

Analytical measurements
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

The apparent quantum yields for the production of BLPs
Spectral apparent quantum yields (AQYs) for the production of BLPs supporting bacterial production (BP) or bacterial respiration (BR) (AQY BP,λ or AQY BP,λ ) were calculated as: AQY BP,λ or AQY BR,λ = BP or BR photons absorbed by CDOM(λ) (1) 5 where BP or BR on BLPs are the bacterial production or respiration based on BLPs (mol C) and photons absorbed represents the spectrally resolved absorption of photons by CDOM during the irradiation experiment (mol photons at the spectral range from 290 to 750 nm) and λ specific wavelength. The latter was calculated by accounting for the spectral dose of photons, the optical path length, and the absorption characteristics of 10 samples as explained in our earlier study (Aarnos et al., 2012). AQY BP,λ and AQY BR,λ were assumed to depend on wavelength according to: where AQY λ is AQY BP,λ or AQY BR,λ , c (mol C mol photons −1 ) and d (nm −1 ) are positive constants, and λ is the wavelength (nm). The parameters c and d in Eq. (2) were iterated from Eq. (1) by unconstrained nonlinear optimization (the "fminsearch" function of the Matlab 7.9.0) as in our earlier publications (Vähätalo et al., 2011;Aarnos et al., 2012).

The production of BLPs based on AQYs
The measured AQY BP,λ or AQY BR,λ was used to calculate the production rate of BLPs 20 in the river plumes according to: 8207 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | where pr is the production rate of BLPs supporting BP or BR over the entire water column (mol C m −2 yr −1 ), AQY λ is AQY BP,λ or AQY BR,λ , Q λ represents the spectrum of solar radiation absorbed by the entire water column (mol photons m −2 yr −1 nm −1 ), and the ratio a CDOM,λ a −1 tot,λ (dimensionless) is the contribution of tDOC to the total absorption coefficient a tot,λ (m −1 nm −1 ) (Aarnos et al., 2012). The integration was done from 5 λ min of 290 nm to λ max of 750 nm. Q λ was calculated as the product of a standard solar radiation spectrum (ASTM G173-03; Chu and Liu, 2009) normalized with global radiation and the annual mean global radiation determined for the area of each river plume examined. The latter was calculated from monthly values over the years 1984 to 2000 (Hatzianastassiou et al., 2005). Terrestrial CDOM was assumed to absorb all pho-10 tolytic photons (a CDOM,λ a −1 tot,λ = 1; Eq. 3). Therefore the calculated rates apply for those coastal regions, where tCDOM from rivers dominates the absorption of photolytic solar radiation (e.g. river plumes).

The production of BLPs based on photobleaching
The annual production of BLPs from tCDOM in the coastal ocean was estimated from 15 the measured relationship between BLPs and photobleaching. The latter was assumed to be the only sink for tCDOM discharged by rivers to coastal ocean (Nelson and Siegel, 2013). In this case, the amount of photobleached tCDOM equals the CDOM flux by the rivers. The annual CDOM fluxes for Lena, Mississippi and St. Lawrence rivers were taken from the literature (Stedmon et al., 2011;Spencer et al., 2013) and 20 converted to a CDOM,300 flux with spectral slope coefficients (Stedmon et al. (2011); 17.6 and 16.3 µm −1 ) measured for Mississippi and St. Lawrence CDOM, respectively.
In the absence of published fluxes for the remaining rivers, the annual CDOM fluxes were estimated by multiplying the measured a CDOM,300 with the annual water discharge (Cauwet, 2002;Coynel et al., 2005). The production of BLPs supporting BP by each 25 river (mol C yr −1 ) was calculated by multiplying the CDOM flux of each river (m 2 yr −1 ) with the relationship (mol C m −2 ) between the BP on BLPs and photobleaching of CDOM at 300 nm. The BR on BLPs was calculated from the mean BGE on BLPs (BR = BP/BGE − BP). Finally, the total production of BLPs (mol C yr −1 ) was obtained as a sum of BR and BP on BLPs.

Statistical analysis
Statistical differences between the samples were tested with two-tailed paired t tests 5 (heteroscedastic) by comparing observed response between the irradiated and the dark control samples. All statistical analyses were carried out by using R language and a significance level of 0.05 (R Core Team, 2014). We used linear regression to evaluate the dependency between photobleaching and and BP or BLPs.

Photobleaching of riverine CDOM in a seawater matrix
When 0.2 µm filtered river waters without labile DOC (Table 1) were mixed with artificial seawater and irradiated, the simulated solar radiation photobleached CDOM at 300 nm (a CDOM,300 ) by 50.4 ± 2.6 % (mean ± sd here and elsewhere in the results; Table 2). The amount of photobleaching expressed as absorption coefficient ranged from 15 0.48 ± 0.06 m −1 (Ganges-Brahmaputra) to 12.42 ± 0.06 m −1 (Congo) when calculated as the difference in a CDOM,300 between the dark control and the irradiated samples (∆a CDOM,300 , Table 2).

Bacterial production
In order to assess bacterial production (BP) on the irradiated and the dark control non-20 labile DOC, we introduced an inoculum of riverine bacterioplankton to the irradiated and the dark control samples resulting in the bacterial cell densities from 0.006 to 0.240 × 10 9 cells L −1 (t = 0 in Table 3). The bacteria increased their densities by one to 8209 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | two orders of magnitude in five to 12 days (Table 3) in the irradiated samples and 33 to 152 % higher more than in the dark control samples (Table 3). The bacterial cell densities were converted to biomasses by accounting for the bacterial cell volumes, which were not different between the dark control and the irradiated samples (paired t test, df = 9, p > 0.05). In the dark controls, the accumulated BP 5 ranged from 0.5 ± 0.1 (Lena) to 3.9 ± 0.2 (Congo) µmol C L −1 (Fig. 1). BP was higher in the irradiated than in the dark control samples (paired t test, df = 9, p = 0.003) and ranged from 1.2 ± 0.1 (Lena) to 8.7 ± 0.6 (Congo) µmol C L −1 (Fig. 1). BP on BLPs, calculated as the difference between the irradiated and the dark control samples, ranged from 0.6 ± 0.3 (Ganges-Brahmaputra) to 4.8 ± 0.8 (Congo) µmol C L −1 (Fig. 1). The BP 10 on BLPs was significantly related to the photobleaching by a linear regression coefficient of 0.336 mmol C L m −2 (Fig. 2).

Bacterial respiration
In the dark controls, bacterial respiration (BR) accumulated usually linearly with time indicating relatively constant rates of respiration throughout the bioassay (Fig. 3). In 15 the irradiated samples, the kinetics of BR was different and rates typically higher than in the dark control samples (Fig. 3). The BR on BLPs (Fig. 4) was calculated from the difference in respiration between the irradiated and dark control samples (Fig. 3). After a lag-phase, BR on BLPs typically accumulated rapidly, but leveled off at the late phase of bioassay (Fig. 4). The BR on BLPs accumulated by the time of maximum ) with Yangtze BLPs (Fig. 4).

Bacterial growth efficiency
When BGE was calculated from the biomass (Fig. 1) and the respiration gained by the time of maximum biomass (Figs. 3 and 4), it was lowest 3.0 % and highest 27.7 % in 25 the irradiated DOC from Yangtze and Ganges-Brahmaputra, respectively, while in the dark controls BGE ranged from 3.6 % (Yangtze) to 15.3 % (Amazon) ( Table 4). BGE on BLPs ranged from 2.8 % (Yangtze) to 21.5 % (Amazon) ( Table 4). The mean BGEs were not different among the irradiated DOC, the dark control DOC or BLPs (paired t test, df = 7, p > 0.05) (Table 4).

5
When the BP on BLPs (Fig. 1) was divided by the spectrum of photons absorbed during the irradiation, the spectral AQY for the production of BLPs supporting BP (AQY BP,λ ; Eq. 2) was defined by the parameters c and d shown in Table 5. Based on these parameters, the calculated AQY BP ranged from 9.5 to 76 with median of µmol C mol photons −1 at 330 nm, a wavelength of solar radiation responsible for highest production of BLPs 10 (Table 5) (Miller et al., 2002;Vähätalo et al., 2011;Aarnos et al., 2012). The corresponding values for the production of BLPs supporting BR (AQY BR,330 ) ranged from 57 to 1203 with median of 186 µmol C mol photons −1 (Table 5). In the river plumes, the annual production of BLPs supporting BP ranged from 1.0 (Lena) to 33 mmol C m 2 yr −1 (Ganges-Brahmaputra) ( Table 6) when calculated (Eq. 3) 15 with AQY BP,λ (Table 5) and the local annual mean solar irradiance in the coastal ocean at the front of each river. The corresponding production of BLPs supporting BR ranged from 19 (St. Lawrence) to 574 mmol C m −2 yr −1 (Yangtze) ( Table 6). The total production of BLPs ranged from 21 (St Lawrence) to 584 mmol C m −2 yr −1 (Yangtze) when calculated as the sum of BLPs supporting BP and BR (Table 6).

An estimate for the global production of BLPs
The relationship between BP on BLPs and ∆a CDOM,300 (Fig. 2) was multiplied with the annual CDOM fluxes (Table 7) to estimate the photobleaching of CDOM and related production of BLPs from non-labile tDOC of each river in the coastal ocean. The photobleaching of CDOM fluxes was calculated to promote altogether 1.1 ± 0.3 Mt C yr −1 25 BP on BLPs (mean ± 95 % confidence interval; Table 7). The corresponding amount of 8211 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | BR on BLPs was calculated to be 9.5 ± 4.8 Mt C yr −1 when accounting for the 12.0 % BGE on BLPs (Table 7). The sum of BP and BR on BLPs was 10.7 ± 4.5 Mt C yr −1 representing the total production of BLPs (Table 7). This total production of BLPs corresponds to 15 ± 6.8 % of the total DOC flux (69 Mt C yr −1 ) of examined rivers ( Table 7). Assuming that the rivers examined responsible for 28 % of global DOC flux are repre-5 sentative for the remaining 72 % DOC flux to ocean, the estimate for the global coastal production of BLPs from tDOC is 38.0 ± 15.9 Mt C yr −1 supporting 4.1 ± 1.1 Mt C yr −1 BP and 33.9 ± 16.9 Mt C yr −1 BR.

10
Our study provides the first estimate for the production of BLPs from tDOC in the coastal waters in the front of ten major rivers across five continents. The BLPs from tDOC (38.0 ± 15.9 Mt C yr −1 ; this study) contribute little to the production of BLPs in global coastal water estimated earlier (206 Mt C yr −1 ) (Miller et al., 2002). That estimate is nearly as large as global riverine DOC flux to the ocean (246 Mt C yr −1 ) (Cai, ). There much of the earlier estimate must be based on the autochthonous production in marine waters and tells little about the coupling of tDOC to marine waters trough the microbial loop.
The annual production of BLPs from tDOC in coastal waters (38.0 ± 15.9 Mt C yr −1 ) estimated in this study is larger than the global photochemical production of dissolved 20 inorganic carbon (DIC) in inland waters (Koehler et al., 2014). This comparison indicates that although solar radiation transforms a remarkable amount of DOC in lakes and rivers, the majority of photochemical processing of terrestrial and freshwater DOC takes place in coastal waters.
According to the present study, the BGE on BLPs is 12 % and thus the majority of as- 25 similated BLPs will be quickly respired to DIC totaling 33.9 ± 16.9 Mt C yr −1 in the global Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | coastal waters. This amount provides an additional source of DIC to coastal ocean in addition to the riverine flux of DIC (407 Mt C yr −1 ; Cai, 2011). This study estimates that BLPs from tDOC support 4.1 ± 1.1 Mt C yr −1 of BP in the global coastal waters. This BP is not alone a source of carbon to food webs, but is associated with N and P bound in bacterial biomass (Smith and Benner, 2005;Vähätalo et al., 2011). Assuming stoi-5 chiometric amounts of N and P associated to bacterial biomass (mass C : N : P ratio of 17 : 4 : 1) (Goldman et al., 1987), the assimilation of BLPs from terrestrial tDOC transfers 4.5 Mt C yr −1 , 1.06 Mt N yr −1 and 0.26 Mt P yr −1 to the coastal food webs through the grazers of bacterioplankton. 10 In this study, the production of BLPs is related to the photobleaching of CDOM with a R 2 0.88 (Fig. 2). Similar relationships has been found also in earlier studies (Miller and Moran, 1997;Obernosterer and Herndl, 2000;Brinkmann et al., 2003). For example, ∆a CDOM,300 explains extremely well (R 2 > 96 %) of the photoproduction of three individual BLP-compounds: formaldehyde, acetaldehyde and glyoxylate (Kieber et al., 15 1990). The corresponding R 2 is lower (0.67) for the relationship between a group of BPLs (carboxylic acids) and ∆a CDOM,365 determined for 38 Swedish lakes (Bertilsson and Tranvik, 2000). Our R 2 -value (Fig. 2) is higher than the corresponding one for 38 lake water samples (Bertilsson and Tranvik, 2000). This is at first glance surprising since our DOC was collected from ten global rivers covering a wide climatic range and 20 land use not limited for a small region (Sweden) as in an earlier study (Bertilsson and Tranvik, 2000). The relatively high R 2 -values in this study may be explained in part by the same chemical matrix (artificial coastal water) used in our experiments. Additionally, the rivers examined here integrate the variability of DOM across their large catchments. For example, one third of 7838 different chemical compounds assigned 25 by ultrahigh-resolution mass spectral analysis to the water samples examined in this study were shared among all river samples (Wagner et al., 2015). A conclusion from 8213 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | this and many other studies (Kieber et al., 1990;Miller and Moran, 1997;Bertilsson and Tranvik, 2000;Obernosterer and Herndl, 2000;Brinkmann et al., 2003) is that the production of BLPs is quantitatively linked the photobleaching of CDOM, which makes it as a useful proxy for the production of BLPs.

BLPs, photobleaching and CDOM fluxes
Our estimate for the global production of BLPs from terrestrial DOM bases on an as-5 sumption that the riverine CDOM flux will be photobleached completely in the coastal ocean. We used published CDOM fluxes for Mississippi, St. Lawrence and Lena Rivers (Stedmon et al., 2011;Spencer et al., 2013), but for the remaining rivers the CDOM flux was estimated simply by multiplying the measured CDOM with the published water discharge (Table 7). This method provides only rough estimates for CDOM fluxes and 10 causes uncertainty also for our global estimates of BLPs. Therefore an additional data on riverine CDOM fluxes in the future will improve our estimate for CDOM fluxes and related production of BLPs. The majority of global riverine CDOM flux enters to the mix-ing layer of coastal waters, where solar radiation photobleaches CDOM efficiently and leaves not traces of terrestrial CDOM to be found in the surface of open ocean (review 15 by Nelson and Siegel (2013) and references therein). Our assumption for complete photobleaching may not hold entirely for the rivers (such as Lena in this study) dis-charging to Arctic Ocean, which transfers 78 200 Gm 2 yr − 1 of riverine CDOM to North Atlantic (calculated for 300 nm from the slope and flux estimated by Granskog et al., 2012). A part of this CDOM flux is directed to the deep ocean without photobleaching 20 as a part of deep water formation in the North Atlantic (Nelson and Siegel, 2013). Thus our assumption for complete photobleaching is relatively well but not entirely supported by our current estimate for photobleaching being the major sink for riverine CDOM in the coastal ocean. 25 Our AQY BP,330 s (9.5-76 µmol C mol photons −1 AQY BR,330 s (57 to 1204 (µmol C mol photons −1 ) are generally lower than the previously reported 1391 (µmol C mol photons −1 ) for river water and 1090 (µmol C mol photons −1 ) for water draining a salt marsh (Miller et al., 2002), but similar to those for rivers and lakes in Alaska (Cory et al., 2014). In the earlier studies, the AQY BP s reported by Aarnos et al. (2012) are lower than AQY BR s reported by Miller et al. (2002) and Cory 5 et al. (2014). To the best of our knowledge, our study presents the first simultaneous determinations of AQY BP and AQY BR and shows that AQYs for bacterial respiration are larger than for bacterial production.

AQYs for BLPs
In this study, BLPs were consumed by riverine bacteria, which survived at the coastal salinities, increased their biomass by one to two orders of magnitude and reached a de-10 tectable peak in biomass within < 12 days. The bacterial inoculum used by Miller et al. (2002) was thousand-times larger and the time of bioassay longer than in the present study. The bacteria used by Miller et al. (2002) were able to assimilate more BPLs resulting in higher AQYs than in our study. Although our bioassays for the determination of BPLs lasted up to 12 days, the initial cell densities of inoculum were low and reached 15 typical densities found in surface water not until the end of bioassay (Table 3). The BLPs determined in this study likely represent the most bioavailable fraction of BLPs (Kieber and Mopper, 1987). The number of known chemically distinct BLP-compounds was 14 in an early review (Moran and Zepp, 1997), but had increased to 44 in a later review (Vähätalo, 2009). A more recent study using a high resolution of mass spectrometry 20 identified 1835 chemical formulae that were produced by solar radiation-induced photochemical reactions (Rossel et al., 2013).From these formulae 44 % were not observed in a treatment with high microbial activity indicating that the number of BLP-compounds was 807 and that BLPs represent a diverse group of compounds (Rossel et al., 2013). Although, by a definition BLPs are biologically more labile than the original DOC, the 25 biodegradability of individual BLP-compounds likely varies and may form a continuum like with the bulk DOM (Moran and Zepp, 1997;Vähätalo, 2009;Vähätalo et al., 2010;Rossel et al., 2013). Thus, as the present study refers to the most bioavailable fraction of BLPs, the entire pool of BPLs is likely larger than what was measured in this study.

8215
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

BGEs on BLPs and non-labile DOC
Our BGE on BLPs (12.0 ± 7.2 %; Table 4) provides the first direct estimate for the BGE of microbial communities on a group of compounds classified as BLPs. Our values can be compared to BGEs reported for single compounds such as acetic and formic acids (Bertilsson and Tranvik, 1998). BGEs on formic and acetic acids has been 2 and 5 41 %, in one study (Bertilsson and Tranvik, 1998), but 20-27 and 73-80 %, respectively, in another study (Remington et al., 2011). The earlier studies demonstrate that BGE on a single BLP compound can vary and thus it is a challenge to determine BGE separately for each BLP compound and compile them together for an overall BGE on all BLPs. Such a simple overall BGE (12.0 ± 7.2 %) can now be applied on BLPs based 10 on the results of the present study.
Our BGE on DOM in the dark controls (11.5 ± 3.9 %, Table 4) represents an estimate for BGE on non-labile tDOC in coastal waters. This estimate falls to the lower range of hundreds BGE determinations done with freshly collected water representing BGEs primarily on labile DOC (Kroer, 1993;del Giorgio and Cole, 1998;Wikner et al., 1999).

Conclusions
Our study shows that the mineralization and transformation of tDOC will continue in the coastal ocean after the quick consumption of labile tDOC at estuaries (Figs. 1-3, Tables 3 and 4). As the tDOC is transported towards open ocean, the microbes will continue to assimilate non-labile DOC with BGEs around 11.5 % (Table 4; Asmala et al., 2014). The photochemical transformation of tDOC is expected to increase after the turbidity maximum, advection and mixing into optically transparent marine waters. These reactions transform tDOC directly into CO 2 Miller and Zepp (1995);Aarnos et al. (2012) and into BLPs to be assimilated by bacteria with BGEs around 12 % (Table 4). The photochemical production of BLPs in coastal waters can be estimated with two ap-5 proaches. The AQYs can estimate the production of BLPs locally in the rivers plumes at any specified time with known solar irradiance (Tables 5 and 6). Similar estimates can be made by multiplying photobleaching of CDOM (Whitehead et al., 2000;Osburn et al., 2001;Vähätalo and Wetzel, 2004) with 0.335 mmol C m −2 (Fig. 2). Photobleaching also provides a robust estimate for the annual production of BLPs by each river or all rivers (Table 7). Although this estimate does not specify where and when the photochemical transformations take place, it improves integrative synthesis on the fate of tDOC in the coastal ocean and its linkages to marine food webs (Bauer et al., 2013).
Khan (Ganges-Brahmaputra), Q. Huang (Yangtze), W. Schneider (Lena), A. Rivas and C. E. Rezende (Amazon), E. Petrescu (Danube), and H. E. Reader (Mississippi). We are grateful to Nicolas Hatzianastassiou for providing the annual doses of global radiation for the river plumes. This research was funded by the Finnish Academy and by the Nordic Center of Excellence CRAICC (Cryosphere-atmosphere interactions in a changing Arctic climate). We also thank 20 the Academy of Finland Center of Excellence (project number 1118615) and Norwegian University of Science and Technology (NTNU).