This article was originally published in a journal published byElsevier, and the attached copy is provided by Elsevier for the

author’s benefit and for the benefit of the author’s institution, fornon-commercial research and educational use including without

limitation use in instruction at your institution, sending it to specificcolleagues that you know, and providing a copy to your institution’s

administrator.

All other uses, reproduction and distribution, including withoutlimitation commercial reprints, selling or licensing copies or access,

or posting on open internet sites, your personal or institution’swebsite or repository, are prohibited. For exceptions, permission

may be sought for such use through Elsevier’s permissions site at:

http://www.elsevier.com/locate/permissionusematerial

Autho

r's

pers

onal

co

py

Photochemical production of ammonium and transformationof dissolved organic matter in the Baltic Sea

Colin A. Stedmon a,⁎, Stiig Markager a, Lars Tranvik b, Leif Kronberg c,Tove Slätis c, Winnie Martinsen a

a Department of Marine Ecology, National Environmental Research Institute, University of Aarhus, Frederiksborgvej 399, 4000 Roskilde, Denmarkb Department of Limnology, Evolutionary Biology Centre, Uppsala University, 752 36 Uppsala, Sweden

c Department of Organic Chemistry, Åbo Akademi University, Biskopsgatan 8, FIN-20500 Turku/Åbo, Finland

Received 9 August 2006; received in revised form 17 November 2006; accepted 29 November 2006Available online 23 January 2007

Abstract

The release of ammonium from the photochemical degradation of dissolved organic matter (DOM) has been proposed by earlierstudies as a potentially important remineralisation pathway for refractory organic nitrogen. In this study the photochemicalproduction of ammonium from Baltic Sea DOM was assessed in the laboratory. Filtered samples from the Bothnian Bay, the Gulfof Finland and the Arkona Sea were exposed to UVA light at environmentally relevant levels, and the developments in ammoniumconcentrations, light absorption, fluorescence and molecular size distribution were followed. The exposures resulted in a decreasein DOM absorption and loss of the larger sized fraction of DOM. Analysis of the fluorescence properties of DOM using parallelfactor analysis (PARAFAC) identified 6 independent components. Five components decreased in intensity as a result of the UVAexposures. One component was produced as a result of the exposures and represents labile photoproducts derived from terrestrialDOM. The characteristics of DOM in samples from the Bothnian Bay and Gulf of Finland were similar and dominated byterrestrially derived material. The DOM from the Arkona Sea was more autochthonous in character. Photoammonification differeddepending on the composition of DOM. Calculated photoammonification rates in surface waters varied between 121 and 382 μmolNH4

+ L−1 d−1. Estimated areal daily production rates ranged between 37 and 237 μmol NH4+ m−2 d−1, which are comparable to

atmospheric deposition rates and suggest that photochemical remineralisation of organic nitrogen may be a significant source ofbioavailable nitrogen to surface waters during summer months with high irradiance and low inorganic nitrogen concentrations.© 2006 Elsevier B.V. All rights reserved.

Keywords: Coloured dissolved organic matter; Ammonium; Photochemistry; Baltic Sea

1. Introduction

It is estimated that globally 70% of the nitrogen ex-ported to coastal waters is bound within the pool of dis-solved organic matter (DOM) (Meybeck, 1982). Free and

combined amino acids, urea, nucleotides, moieties offulvic and humic acids and other uncharacterised highmolecular weight biomolecules all contribute to the poolof dissolved organic nitrogen (DON) (Bronk, 2002). Thebulk of DON is present in the largely uncharacterisedfraction (Bronk, 2002; Benner, 2002). As a result manyaspects of the role of dissolved organic nitrogen as anutrient source for marine production are not fully

Marine Chemistry 104 (2007) 227–240www.elsevier.com/locate/marchem

⁎ Corresponding author. Tel.: +45 4630 1805; fax: +45 4630 1211.E-mail address: cst@dmu.dk (C.A. Stedmon).

0304-4203/$ - see front matter © 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.marchem.2006.11.005

Autho

r's

pers

onal

co

py

understood. One such process is photoammonification;the production of ammonium upon the exposure of DOMto UV light (Bushaw et al., 1996).

Over the last 10 years a series of studies on theoccurrence and significance of photoammonificationhave been carried out. The results published appear to becontrasting (see brief review in Grzybowski (2003)).They can be classed into four categories; i) detectableammonium production at environmentally relevant rates(e.g. Bushaw et al., 1996; Gao and Zepp, 1998; Vähätaloet al., 2003), ii) detectable ammonium production withquestionable ecosystem relevance (Buffam andMcGlathery, 2003), iii) no detectable production (e.g.Jørgensen et al., 1998; Bertilsson et al., 1999; Wiegnerand Seitzinger, 2001) and iv) photochemically mediatedincorporation of ammonium into DON (Kieber et al.,1997; Koopmans and Bronk, 2002; Vähätalo et al.,2003). These results likely reflect the fact that the processis specific to the original composition of the DOM pool,which can be expected to vary both geographically andseasonally. Additionally ambient concentrations of am-monium appear to influence the net result with highammonium concentrations tending to result in a net lossof ammonium as a result of photodegradation (Kieberet al., 1997; Koopmans and Bronk, 2002).

Before comparing photochemical production rates orquantities between studies or samples it is important tonormalise the production to the amount of light actuallyabsorbed by the sample in the reaction vessel (Hu et al.,2002; Vähätalo and Zepp, 2005). The experimentalconditions (e.g. intensity, spectral properties of theirradiation, path length and shape of incubation vessels,and exposure time) influence the result and make inter-study or even inter-sample comparison often difficult(Hu et al., 2002).

As with many coastal seas in densely populatedregions, the Baltic Sea suffers from eutrophication. It hasa limited water exchange with the Atlantic Ocean (via theNorth Sea) resulting in riverine inputs having consider-able influence on its nutrient budgets. Riverine inputs areestimated to be responsible for 60% of the nitrogensupply to the Baltic (Stålnacke et al., 1999). Otherimportant sources are atmospheric deposition andphytoplankton nitrogen fixation. Although about halfof the freshwater total nitrogen supply is as dissolvedinorganic nitrogen (DIN) (Stepanauskas et al., 2002),estuarine and near-shore biological productivity trans-forms a substantial amount DIN to DON (Stedmon et al.,2006) before it is transported offshore to the open BalticSea. This exported DON is a combination of terrestriallyderived material and coastal/estuarine autochtonouslyproduced material, and little is known about its

susceptibility to photochemical degradation. Recently astudy in the Gulf of Finland showed that DON can bephotochemically mineralised to ammonium using sim-ulated sunlight (Vähätalo and Zepp, 2005). Photo-ammonification was found to occur at rates which weresimilar to atmospheric deposition and thought to be animportant source of bioavailable N, in offshore waters.

The aim of this work was to study the photochemicalproduction of ammonium from samples from differingsites in the Baltic Sea and assess the its significance as anitrogen source for offshore productivity. In conjunctionchanges in the optical characteristics (absorption andfluorescence properties) and molecular size distributionof the DOM were followed in order to detect the effectsof photodegradation on DOM characteristics and takeinto account qualitative and quantitative differencesbetween samples before and during the exposures.

2. Methods

2.1. Sampling and experimental set up

Water samples were collected during summer 2005from three regional water quality monitoring stations inthe Baltic Sea; Arkona Sea (St444, 55°00′N, 13°18′E,August), Kotka (St12370, 60°15′N, 27°15′E, August),and Oulu (St30372, 65°08′N, 24°36′E, September)(Fig. 1). Surface water samples were collected andplaced in 5 L acid washed polyethylene containers andkept refrigerated (4 °C) and in the dark for 2–3 daysbefore arriving at NERI, Denmark. In the laboratoryeach sample was first filtered through a pre-combustedGF/F filter, then subsequently filtered through 0.2 μmMillipore filter (pre-cleaned with 500 mL of Milli-Q) inorder to remove the majority of bacteria. After filtrationthe water from each site was distributed into 24 quartztubes (approximate volume of 200 mL). Twelve of thetubes were wrapped in aluminium foil and used as a darkcontrol, whilst the remaining twelve were placed in frontof a UV-A lamp (Q-panel, UVA340). The dark controlswere incubated next to the exposed samples. There wasno headspace in the tubes, and they were sealed withground glass stoppers. Incubations were carried out atroom temperature and the experimental set-up was blackin order to limit internal reflection of light. Fig. 2 showsthe spectral distribution and intensity of the lampirradiance incident on the quartz tubes. The spectralproperties of the lamp light and its intensity in the UVregion of the spectrum was equivalent to that observedin summer at these latitudes (Fig. 2). Triplicate tubesfrom each treatment (UVA and dark) were harvestedafter 0, 6, 24 and 48 h. As an additional control for

228 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

possible contamination during the incubations 12 tubeswere also filled with Milli-Q water, placed in front of theUVA lamp and sampled at the same time intervals. Ateach time interval concentrations of ammonium (NH4

+),and the molecular size distribution and optical char-acteristics of DOM (spectral fluorescence and absorp-tion) were measured.

2.2. Ammonium measurements

NH4+ measurements were made following the fluoro-

metric technique described by Holmes et al. (1999). Theapproach uses a working reagent consisting of ortho-phathaldialdehyde (OPA, Sigma P-1378), sodium sulfite(Sigma S-4672) and sodium tetraborate (Sigma S-9640).

For the experiment a∼2 L batch of working reagent wasmade by mixing 100 mL of an OPA solution (4 g in100 mL of high grade ethanol), 2 L of a sodiumtetraborate solution (80 g in 2 L of Milli-Q water) and10 mL of a sodium sulfite solution (1 g in 125 mL Milli-Q water). For sampling, 40 mL of sample water wasplaced in an acid washed 100 mL amber glass bottle and10 mL of working reagent was added. The samples werethen subsequently shaken and allowed to stand at roomtemperature and in the dark for 2–3 h. A series of 6standards were made using an ammonium stock solution(50 μmol L−1) to 40 mL of Milli-Q water in acid washed100 mL amber glass bottles. Standards were made ateach sampling and treated in the same way as samples.All standard curves were linear with a r2N0.999.

Fig. 1. Map of the Baltic Sea with the locations of the three sampling stations marked.

Fig. 2. Comparison of the spectral quality of the Q-panel UVA340 light source with a sunlight spectrum measured in Denmark (midday in July). Bothmeasurements were made using a Licor 1800 spectroradiometer.

229C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

The fluorescence of the samples and the standardswere measured on a Varian Eclipse fluorescencespectrophotometer with an excitation wavelength of360 nm and an emission wavelength of 425 nm. Both theexcitation and emission slit widths were set to 5 nm.Background fluorescence (fluorescence not due to thepresence of NH4

+) was monitored by measuring thefluorescence in samples where no working reagent wasadded. No significant background fluorescence wasdetected with this instrument set up (low PMT voltage).The fluorescence of dissolved organic matter, which isresponsible for the majority of background fluorescenceassociated with this technique, had a fluorescence signalwhich was more than an order of magnitude lower thanthe OPA-NH4

+ signal. In addition matrix effects weredetermined and accounted for by comparing the effectsof standard addition to a Milli-Q water and sample water(Holmes et al., 1999).

2.3. Size exclusion chromatography

High performance size exclusion chromatography(HPSEC) analyses were carried out on the DOM samplesaccording to the technique described in Peuravuori andPihlaja (1997). Measurements were performed with anAgilent 1100 Series liquid chromatographic systemconsisting of a quaternary pump, a vacuum degasser,an autosampler, a thermostated column compartment, avariable wavelength UV-detector (operating at 254 nm)and Agilent ChemStation data system (Agilent Technol-ogies, Espoo, Finland). The separations were performedat 30 °C on a silica based TSKgel G3000SW column(7.8 mm ID×300 mm) and the column was elutedisocratically with 10 mmol L−1 ammonium acetate and1% isopropanol at a flow rate of 1 mL min−1. Theinjection volume was 100 μL.

2.4. Optical analysis

Spectral absorbance (240–700 nm) of coloureddissolved organic matter (CDOM) was measured ona Shimadzu UV-2401PC spectrophotometer using5 cm quartz cuvettes and with Milli-Q water as areference. The absorption coefficient was obtainedby Eq. (1).

aðkÞ ¼ 2:303AðkÞ=L ð1Þ

Where aλ is the absorption coefficient (m−1) at awavelength λ, and A is the absorbance measured and Lis the cuvette length in meters. The absorption spectrawere modelled according to Eq. (2) over wavelengths

from 300–650 nm using the approach described inStedmon and Markager (2001).

aðkÞ ¼ aðk0Þe−Sðk0−kÞ þ K ð2Þ

λ0 is 400 nm and K is a background constant thatallows for any baseline shifts. The coefficient S is thespectral slope coefficient and is a measure of theapproximate exponential decline in absorption withincreasing wavelength.

Spectral fluorescence measurements were made on aVarian Eclipse fluorescence spectrophotometer. A seriesof emission spectra from 300–600 nm (every 2 nm) wererecorded whilst sequentially exciting at wavelengthsfrom 240 to 450 nm every 5 nm. The excitation andemission slit widths were set to 5 nm. The fluorescencespectra measured from each sample were combined togive excitation emission matrices (EEMs). EEMs wereRaman calibrated (normalised to the integral of theRaman band from excitation at 350) to the signal from aMilli-Q water sample run the same day as the samples(Stedmon et al., 2003). In addition the EEMs were cor-rected for instrument specific spectral bias and sampleinner filter effects (Mobed et al., 1996; Lakowicz, 1983;McKnight et al., 2001). Excitation correction spectrawere derived using Rhodamine B and emission correc-tion using a ground quartz diffuser (Varian Inc.).

2.5. PARAFAC analysis

The EEMs were subsequently characterised by par-allel factor analysis (PARAFAC) in MATLAB softwareusing the “N-way toolbox for MATLAB” (Bro, 1997;Andersson and Bro, 2000; Stedmon et al., 2003).PARAFAC is an alternating least squares regressionprocedure that splits the EEMs into a series of tri-linearterms and a residual array.

xijk ¼XFf¼1

aif bjf ckf þ eijk ; i ¼ 1; N ; I ; j

¼ 1; N ; J ; k ¼ 1; N ;K ð3Þ

When PARAFAC is applied to fluorescence EEMsxijk represents the fluorescence intensity for sample i atemission wavelength j and excitation wavelength k. aifis directly proportional to the concentration (e.g. μM C)analyte f in sample i. bjf is directly proportional to thequantum efficiency of fluorescence of analyte f atemission wavelength j. Likewise, ckf is linearly relatedto the specific absorption coefficient at excitationwavelength k. F is the number of components in the

230 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

model and εijk is the residual matrix (unexplainedsignal). This approach is an iterative process whichseparates the signal from a complex mixture into itsunderlying components, with no assumptions on thenumber of components or their spectral shape.

All the EEMs measured were pooled into a dataarray. A total of 72 EEMs were measured. Visualinspection of the triplicate samples and assessment ofinitial PARAFAC models identified five erroneoussamples (EMMs) and they were therefore excludedfrom the final PARAFAC analysis. The data arrayconsisted of 67 samples by 151 emission wavelengthsby 41 excitation wavelengths. The data array was thensplit into two halves and the PARAFAC algorithm wasrun stepwise on both halves, for 3–10 components. Theappropriate number of components was determined by asplit-half validation and residual analysis (Bro, 1997;Stedmon et al., 2003; Stedmon and Markager, 2005a).

2.6. Photochemical calculations

The formula used for estimating the photoproductionof ammonia (ηNH4

) in the quartz tubes is,

gNH4ðl mol m−3s−1Þ

¼Z kmax

kmin

E0;kϕNH4;kaCDOM;k½1−expðat;kLÞ�=ðat;kLÞdk

ð4Þ(Hu et al., 2002; Vähätalo and Zepp, 2005).E0,λ (μmol

photons m−2 s−1 nm−1) is the spectral irradiance of theincident light at the surface of the quartz tubes, uNH4,λ

(mol NH4+ (mol absorbed photons)−1) is the apparent

quantum yield for photoammonification, aCDOM,λ (m−1)

is CDOMs absorption coefficient, is the total lightabsorption of the sample (sum of aCDOM,λ and awater,λ(absorption by water obtained from Smith and Baker(1981)), and L (m) is the optical path length.

In this study the geometric mean of at the start andend of the exposure period (Eq. (4)) and was used torepresent CDOM absorption between two samplingtimes (Miller and Zepp, 1995).

agmCDOM;kðm−1Þ ¼ 10ððlogðat1CDOM;kÞþlogðat2CDOM;kÞÞ=2Þ: ð5Þ

The optical path length in the quartz tubes used wasapproximated by Eq. (6) (Vähätalo and Zepp, 2005)

L ¼ 0:5krl−1k : ð6Þ

Where r is the internal radius of tubes (in this study0.0185 m) and μλ is the mean cosine of photons in the

tubes. μλ was not determined for this experiment how-ever can theoretically vary between 0.5 and 1 (Vähätaloand Zepp, 2005). For the estimations of uNH4,λ in thisstudy three values of μλ were used (0.5, 0.75 and 1) inorder to reveal its influence on estimates.

The apparent quantum yield spectrum photoammo-nification can be assumed to behave exponentially andbe approximated by Eq. (7) (Vähätalo and Zepp, 2005).

ϕNH4;k ¼ cexpð−dkÞ ð7Þ

Estimates of the constants c and d were obtained bycombining Eqs. (4) and (7) and applying an iterative leastsquares optimisation routine to spectral data from 300 to450 nm (total of 151 wavelengths), using the “fmin-search” function in Matlab7 software (Vähätalo andZepp, 2005). Values of ηNH4

used for the optimisationwere calculated by the difference between the exposedand dark ammonium concentrations after 48 h, normal-ised to the incubation time in seconds.

3. Results

3.1. The optical and chemical DOM characteristics

There were clear differences in the chemical andoptical characteristics of the DOM from the three sam-pling sites (Table 1). The greatest difference was be-tween the Kotka and Arkona samples, whilst the samplefrom Oulu had characteristics that were intermediate.The DOM in the Kotka sample resembled that of terres-trially derived material with a high DOC concentrationand a350, a spectral slope (S) value of 18.5 μm−1, and aDOC:DON ratio of ∼30. In contrast the sample fromArkona had a lower DOC and a350 values, a high S value(23.9 μm−1) and a lower DOC:DON ratio. The majorityof dissolved N in the samples was as DON. In the Arkona

Table 1Optical and chemical characteristics of the samples before theincubations

Salinity DOC(μmolL−1)

a350(m−1)

S(μm−1)

TDN(μmolL−1)

DON(μmolL−1)

DOC:DON

Arkona 7.7 316 1.49 23.9 18.5 10.2 17.1Kotka 3.2 574 6.60 18.5 18.9 18.0 30.4Oulu 3.0 347 3.33 19.8 12.5 7.8 27.8

a350 is CDOM absorption at 350 nm and S is the spectral slope (Eq.(2)). DOC (Dissolved organic carbon), TDN (total dissolvednitrogen) and DON (dissolved organic nitrogen) determinedaccording to techniques described in Stedmon et al. (2006).DOC:DON is the molar ratio of DOC to DON.

231C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

sample approximately 55% of the total dissolved N wasorganic while in the Kotka sample the value was 95%.

The fluorescence-PARAFAC analysis identified 6fluorescent groups (components). Their spectral proper-

ties are shown in Fig. 3. The samples from each sitediffered quantitatively and qualitatively in their fluores-cence characteristics (Fig. 3). The greatest fluorescenceintensity was in the Kotka sample where the fluorescence

Fig. 4. Size exclusion chromatograms of CDOM from a) Arkona, b) Oulu and c) Kotka at the start of the experiment. One of the chromatograms isannotated to show the peak areas integrated and discussed in the Results section.

Fig. 3. The spectral properties of the six fluorescent components identified by the PARAFAC analysis. The bar graph depicts the fluorescence of eachcomponent in each of the three original samples (before exposure). The values are averages of triplicate samples. Error bars represent the standarddeviations. The fluorescence is at the excitation and emission maximum for each component and in Raman units (nm−1).

232 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

Fig. 5. Development in the fluorescence of all six components during the incubations. Data represent the mean of triplicate samples and error bars areone standard deviation. Where error bars are not apparent they are smaller than the size of the symbol. The fluorescence is at the excitation andemission maximum for each component and in Raman units (nm−1).

233C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

pyof components 1 and 2 was more than twice of that in theArkona and Oulu samples. The fluorescence of DOMfrom Kotka and Oulu was similar in composition anddominated by components 1 and 2, whilst the fluores-cence of components 3, 4, 5 and 6 were relatively minor(Fig. 3). Component 6 was very low in the DOM fromOulu (0.02±0.002 nm−1), however, more prominent inthe sample fromKotka (0.082±0.014 nm−1). TheArkonasample differed most notably from the others in that itsfluorescence was dominated by component 4 (Fig. 3).

The size exclusion chromatograms of the samples areshown in Fig. 4. The general chromatogram profileswere very similar from sample to sample and it waspossible to define 5 specific regions or peaks in thechromatograms. These regions were labelled A, B, C, Dand S and integrated in order to facilitate the furtherinterpretation of the results. Peak A had the lowestretention time in the column and therefore representedthe largest sized material. The integrals of the peaks

A–D varied from sample to sample and the highestpeak intensities were noted for the Kotka sample andthe lowest for the Arkona sample. Further, the CDOMsize distributions varied between locations and espe-cially the Arkona sample differed from the other twosamples. The size distribution of CDOM from Arkonadiffered noticeably from the other two samples. Thepeak integrals for A, B and C were of a similarmagnitude for the Arkona sample whereas in theKotka and Oulu samples peak A dominated, suggest-ing that the DOM was characterised by a greaterproportion of larger sized material (Figs. 4 and 6).

The region marked S was found to be related to thesalinity of the water. Tests in the laboratory revealed thatthe peak is absent in desalinated sea water samples.Additionally it was found that the peak can be generatedupon addition of sodium chloride to fresh water CDOMsamples (data not shown). At present it is unclear whatmaterial the peak represents.

Fig. 6. Effects of the UVA exposure on the integrated peaks in the SEC chromatograms son in Fig. 4. The values are averages of triplicate samples andthe error bars represent one standard deviation. (a.u. = arbitrary units).

234 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

3.2. Photochemical alterations of CDOMscharacteristics

In the Kotka and Oulu samples no significantchanges in CDOM absorption (a350) were measuredduring the dark incubation. For the Arkona sample therewas a small but significant (PN0.01) linear increase ina350 during the dark incubation of approximately

0.0017 m−1 h−1. In contrast, the UVA exposure causedCDOM absorption in all samples to decrease with time.The loss in CDOM absorption differed between sitesand with exposure time. Forty-eight hours exposurereduced a350 values by 38, 49 and 50% for Arkona,Kotka and Oulu samples respectively.

The 48 hour incubations with UVA light had a con-siderable influence on the fluorescence composition andintensity (Fig. 5). The dark control incubations for theArkona and Oulu samples exhibited no change influorescence, whereas in the Kotka samples the fluores-cence decreased significantly in an approximately linearfashion. Between the components there were threedifferent responses to UVA exposure; removal, produc-tion, and a combination of initial production and sub-sequent removal. During the UVA incubations thefluorescence of components 1, 2, 4 and 5 was reducedon average by 65±7%, 69±6%, 54±6%, and 26±9%(mean±standard deviation) respectively, after taking intoaccount the slight dark loss of fluorescence observed intheKotka sample. The fluorescence of component 3 in theArkona sample was initially low and apparently noteffected by the UVA exposure (Fig. 5). In contrast in theEastern Baltic samples the fluorescence of component 3initially increased during the first 6 h of the exposure toUVA, then returned to approximately start values after24 h, and was subsequently removed during the finalstages of the exposure. The UVA incubation resulted in anet production of component 6 in the samples from allthree sites. The fluorescence of this component increasedby 0.022 nm− 1 (252%), 0.028 nm− 1 (34%) and0.043 nm−1 (2740%) respectively, in the Arkona, Kotkaand Oulu samples.

In general there were no significant changes in thesize exclusion peak integrals during the incubation ofthe samples in the dark (Two-way ANOVA, PN0.01)(Fig. 6). The intensity of the peaks representing thelarger size material (regions A and B) decreased sig-nificantly (Pb0.01) during the exposure of DOM fromKotka and Oulu (Fig. 6). For the Arkona samples onlythe integral of peak A decreased significantly (Pb0.01).The intensity of the smaller size fractions, C and D,showed no significant (PN0.01) change during theexposure (Fig. 6).

3.3. Photoammonification

In all three incubations exposure to UVA light resultedin increased NH4

+ concentrations relative to the darkcontrol (Fig. 7). For the Arkona sample the averageconcentration in the dark series was 2.0 μmol NH4

+ L−1,whilst in the UVA series the concentration increased to a

Fig. 7. Ammonium concentrations in the UVA exposed and darkincubations from the three sampling sites. Each measurement is anaverage of three samples (true replicates). Error bars indicate standarddeviations. First sample (t=0) for the Arkona UVA series was lost andtherefore missing. In calculations it has been assigned the value of thedark, t=0 samples.

235C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

maximum of 3.7 μmol NH4+ L−1 after 48 h. For the Kotka

sample, maximum concentrations were measured after24 h of exposure after which there was no furthersignificant production of NH4

+. In the dark series theaverage concentration was 1.4 μmol NH4

+ L−1 whilst theconcentration after 48 hour exposure to UVA light was2.9 μmol NH4

+ L−1. There was no NH4+ production in the

UVA series for the sample from Oulu. Concentrationsremained around 0.96 μmol NH4

+ L−1 (standard deviation0.1 μmol NH4

+ L−1). However in the dark series theconcentration decreased by a factor of two from0.94 μmol NH4

+ L−1 at the start to 0.48 μmol NH4+ L−1

after 48 h. Ammonium concentrations measured in quartztubes containing Milli-Q water were very low did notincrease during the incubation, hereby demonstrating thecleanliness of the incubation procedure (Fig. 5A).

The parameters describing the apparent quantum yieldspectra for ammonium production (c and d in Eq. (7)) arepresented in Table 2. As μλ was not determined for thisexperimental set-up the spectra were estimated fordiffering values of μλ across its theoretical range from0.5–1. The results show that the different values ofμλ hada limited effect on the estimates of c and d. From hereonwards the results from using a μλ of 0.75 will bediscussed.uNH4,350 varied between 14 and 83×10

−6 withthe greatest values found for theArkona sample. The dailyrates of photoammonification in surface waters can beapproximated by integrating the product of the dailyphoton flux (mol m−2 d−1 nm−1), CDOM absorptionspectrum (m−1) and the apparent quantum yield ofammonification (Vähätalo and Zepp, 2005). This wasdone using a typical Danish summer irradiance spectrum(Fig. 8a), and integrating over wavelengths from 300–600 nm. Surface photoammonification rates were 321,382 and 121 μmol NH4

+ L−1 d−1 for Arkona, Kotka andOulu samples respectively. Additionally the areal spectralphotoammonification rates were calculated as the productof the daily irradiance (mol m−2 d−1 nm−1) and theapparent quantum yield of ammonification (Fig. 8b).Maximum values occurred at 330 nm. The integratedareal photoammonification rates were greatest for theArkona sample with a value of 237 μmol NH4

+ L−1 d−1.The Kotka and Oulu samples had rates of 60 and 37 μmolNH4

+ L−1 d−1 respectively.

Table 2Apparent quantum yield (AQY) parameters for photoammonification determined between 300 and 450 nm and the AQY at 350 (uNH4,350)

Sample Period(hours)

μλ=0.5 μλ=0.75 μλ=1

c d (nm−1) uNH4,350 (×106) c d (nm−1) uNH4,350 (×10

6) c d (nm−1) uNH4,350 (×106)

Arkona 0−48 1.11 0.0272 81 1.10 0.0271 83 1.09 0.0271 83Kotka 0−48 1.01 0.0306 23 1.03 0.0306 23 1.02 0.0304 24Oulu 0−48 0.99 0.0319 14 0.96 0.0318 14 0.97 0.0317 15

Vähätalo and Zepp (2005)

Mean 0.93 0.0322Min 0.88 0.0296Max 1.06 0.0341

As the mean cosine of photons in the quartz tubes (μλ) was not measured the AQYparameters have been determined for values across its theoreticalrange 0.5–1. Also shown are average values reported for the Gulf of Finland by Vähätalo and Zepp (2005).

Fig. 8. The top panel shows the daily solar radiation spectrum used in thecalculations and the bottom panel shows the calculated areal spectralphotoammonification for the three samples. Also shown is the spectrumderived from data in Vähätalo and Zepp (2005) for the easterly station inthe Gulf of Finland (close proximity to the Kotka station in this study),using an a300=9.9 m−1, S=18.8 μm−1, c=1.06, d=0.0296 nm−1.

236 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

4. Discussion

The composition of DOM in the Baltic Sea can beexpected to vary depending on location and season.Proximity to rivers will result in the DOM resemblingthat of terrestrially derived material. Terrestrial DOM isgenerally characterised by being relatively N poor (highDOC:DON ratio) and having a large fraction of highmolecular weight material (Benner, 2002; Bronk, 2002;Stepanauskas et al., 2002). Additionally the opticalproperties of terrestrial DOM are characterised byspectral slope values between 18 and 19 μm−1 and afluorescence signal that is dominated by broad fluores-cence peaks at wavelength greater than 400 nm(commonly referred to as humic-like fluorescence)(Coble, 1996; Stedmon and Markager, 2001; Kowalc-zuk et al., 2006). Dilution and degradation of riverineDOM in combination with autochthonous DOMproduction in offshore waters results in a change inthe general characteristics. The relative nitrogen contentincreases and the average molecular weight decreases.Similarly the optical properties hereby change withspectral slope values typically increasing and thefluorescence signal becoming less dominated byhumic-like fluorescence and the fluorescence at wave-lengths below 400 nm (protein-like fluorescence)becoming more prevalent (Coble, 1996; Stedmon andMarkager, 2001; Kowalczuk et al., 2006). The chemicaland optical measurements on the start samples revealedthat there are differences between the characteristics ofDOM from the three areas of the Baltic Sea. The DOMfrom Oulu and Kotka resembled that of terrestriallyderived material whereas the properties of the DOMfrom the Arkona Sea differed, reflecting the greaterautochthonous nature.

4.1. Photochemical alterations of DOMs characteristics

The fluorescence characteristics of components 1, 2and 3 are similar to the humic fluorescence reported inearlier work (Coble, 1996; Blough and Del Vecchio,2002; Cory and McKnight, 2005; Stedmon andMarkager, 2005a,b). In particular, component 3 withemission maximum at longer wavelengths (∼500 nm)has been identified in a variety of environments rangingfrom lakes, streams, estuaries, coastal waters and theopen ocean and is thought to represent terrestriallyderived material (Søndergaard et al., 2003; Cory andMcKnight, 2005; Stedmon et al., 2005a; Murphy et al.,2006). Component 3 is also known to be susceptible toboth microbial and photochemical degradation(Søndergaard et al., 2003; Stedmon and Markager,

2005a) as was the case in this study for the samples fromKotka and Oulu. In both these samples there was a netremoval during UVA exposure. However, during thefirst 6 h of the UVA incubation there was a netphotochemical production of this component. Increasesin fluorescence in this region during irradiations havebeen reported before however not explained (Kouassiand Zika, 1990). The results of this study suggest thatcomponent 3 is an intermediate that is formed as a resultof the photochemical degradation of terrestrial DOM. Itis in itself also susceptible to photodegradation howeverthis is delayed until the concentrations of component 1and 2 are reduced to similar levels, possibly be due toinner filter effects by other light absorbing components(both fluorescent and non-fluorescent). In the Arkonasample no changes were observed, suggesting that theprecursor material was absent and that at theseconcentrations and exposure times, no significantphotochemical degradation of component 3 occurs.

Components 4 and 5 had a fluorescence signal whichis similar to free and protein bound amino acids. Fluo-rescence in this region is often ascribed to autochthonousDOM (Coble, 1996; Cory and McKnight, 2005). Boththese components were removed during the exposure toUVA radiation and are likely candidates for the source ofammonium produced during the experiment. This issupported by the fact that the greatest photoammonifica-tion rates were measured in the Arkona sample whichhad a fluorescence signature dominated by component 4and a low DOC:DON ratio. However no direct linearcorrelation was found between ammonium productionand the fluorescence of these components.

Component 6 was found to be produced photochem-ically. A recent study by Murphy et al. (2006) alsoidentified a fluorescent component in coastal and oce-anic samples, with characteristics very similar to ourcomponent 6. These results suggest that this componentis a common product of photodegradation in a range ofenvironments. However, a recent study on the produc-tion and fate of autochthonous fluorescent DOM in themarine environment did not identify this componentdespite the similar equipment and techniques beingapplied (Stedmon and Markager, 2005b). This indicatesthat component 6 may be a photoproduct of terrestri-ally derived DOM only. The fact that its fluorescence islow in the waters flowing out of the Baltic (Arkona data,Fig. 3) suggests that it is also being removed eithermicrobially and/or physically during its passage throughthe Baltic Sea. A range of earlier studies have shown thatphotodegradation of terrestrial DOM can result in theproduction of labile low molecular weight compounds(Moran and Zepp, 1997; Bertilsson et al., 1999; Vähätalo

237C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

et al., 2003) and it is possible that component 6fluorescence represents this material, however additionalexperiments are required before this can be confirmed.

The fluorescence analysis revealed that Baltic SeaDOM consists primarily of 6 different fluorescent groupswhich differ in relative contribution depending onlocation (source of DOM). Photodegradation influencedthese components in a similar manner independent ofsampling site. The results revealed that exposure of theDOM from Oulu and Kotka caused the humic fluores-cence (component 1–3) to decrease to similar levels tothat found in the Arkona Sea. A similar trend was ob-served in the size distribution chromatograms. Afterexposure to 48 h of UVA light, the size distribution of theCDOM from Kotka and Oulu resembled that fromArkona (Fig. 7). These results suggest that thephotochemical experiment has mimicked the transfor-mations occurring within the CDOM pool in Balticsurface waters during its passage westwards. With time,the terrestrial CDOM supplied to the Eastern Baltic Seaby rivers is degraded photochemically resulting in ageneral decrease in the average molecular size.

4.2. Photoammonification rates in the Baltic Sea

For the samples from Arkona and Kotka the con-centrations of ammonium increased in the exposed tubeswhile remaining constant in the dark tubes, as wasexpected. However, for the Oulu sample little or noproduction was observed in the exposed tubes and therewas a loss of ammonium in the dark control series. Theexact reason for this is unknown due to the lack ofauxiliary measurements, however there are several pos-sibilities. The results could suggest that the sterile fil-tration was not successful and there was bacterial growthduring the dark incubations, consequently removingammonium. In the UVA exposed series, bacterial growthwould have been greatly hindered due to the damagingeffect of UV light on bacteria explaining why little or nochange in ammonium was observed. An alternativeexplanation for the decline in ammonium concentrationsmay be that the free ammonium in the Oulu sample wasgradually adsorbed to the humic material. Humics arethought to be capable of loosely binding ammonium andreleasing it again via cation exchange upon increases inionic strength (Rosenfeld, 1979; See, 2003). The lack ofnoteworthy change in ammonium concentrations in theUVA exposed series could therefore be explained byeither the photochemical alteration of the ion exchangesites or a balance between photochemical release andhumic adsorption of ammonium. A prerequisite for thisexplanation to the ammonium decrease is that the

ammonium adsorbed to humic matter is withdrawnfrom detection by the ammonium assay used in thisstudy. Dark adsorption of ammonium by humic materialhas been reported before for DOM from three rivers inthe South East USA (See, 2003).

The total production of a photoproduct is dependenton the number and/or energy (wavelength) of photonsabsorbed. In order to accurately compare the photo-bleaching of CDOM and any subsequent ammoniumproduction between samples the results first have to benormalised to the experimental conditions (Hu et al.,2002). The number of photons absorbed is dependent oni) the absorption properties of the sample, ii) spectralproperties and intensity of incident light, iii) length ofexposure period, and iv) dimensions of the reactionvessel (i.e. path length and cross sectional exposed areaof the quartz tubes). Many studies in the literature havenot taken these factors into account (see Hu et al. (2002)for a summary) and this makes it difficult to compareresults or identify consistent trends (Grzybowski, 2003).

A recent study by Vähätalo and Zepp (2005) de-scribed a technique to determine the apparent quantumyield spectra for photoammonification. The same ap-proach was applied here and similar parameter estimateswere obtained (Table 2). The previous study reportedvalues between 0.88−1.06 and 0.0296–0.0341 nm−1 forc and d respectively at stations in the Gulf of Finland.The values obtained for the Kotka sample in this studyfall within these ranges (Table 2). The consistencybetween the estimates from Kotka in this study andearlier estimates, despite the differing experimental set-up and sampling years, suggests that there may belimited variability in the apparent quantum yield ofphotoammonification within a region. However seasonalpatterns have yet to be investigated.

The apparent quantum yield spectra for Oulu andKotka were similar which is also to be expected as theDOM chemical and optical characteristics were alsoalike, indicating a similar composition (Table 1; (Figs. 3,4 and 6)). The Arkona sample differed with a slightlyhigher estimate for c and lower estimate for d. Thisperhaps reflected the fact that its chemical character-istics also deviated with the DOM have a greaterautochthonous character.

Ammonium concentrations in Baltic offshore surfacewaters typically vary between undetectable and∼2 μmolL−1 depending on remineralisation rates and phytoplank-ton productivity (Grasshoff and Voipio, 1981). Originalammonium concentrations at the start of the experimentwere at similar levels to that that could be producedphotochemically (0.8–2.2 μmol L−1). The areal photo-ammonification rates ranged from 37 μmol N m−2 d−1 in

238 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

the Bothnian Bay to 237 μmol N m−2 d−1 in the ArkonaSea. The value of 60 μmol N m−2 d−1 for the Kotkasample agrees well with the estimate of 53 μmol N m−2

d−1 derived by Vähätalo and Zepp (2005). For compar-ison the daily atmospheric nitrogen deposition in theBaltic (calculated from annual rates) have been estimatedto vary between 40 μmol N m−2 d−1 to 196 μmol N m−2

d−1 depending on location, with the highest valuesreported in the southern Baltic (Hertel et al., 2003). Thesevalues are of similar magnitude to the photoammonifica-tion results in this study and suggest that photoammoni-ficationmay provide an important, as yet unaccounted for,source of available nitrogen in the Baltic. Although theserates are not high enough to establish phytoplanktonblooms (Carstensen et al., 2005) they are likely to be animportant source of nitrogen during warm dry summermonths with limited atmospheric deposition and nitrogenlimited surface water productivity.

In conclusion the results of this study show thatphotoammonification rates vary depending on thecomposition of DOM. Daily areal production rates are afactor of 4–6 times greater at sites where DOM is moreinfluenced by autochthonous production. In offshore wa-ters during the summer months DOM is often dominatedby autochthonous material and pelagic productivity isnitrogen limited. Under fair weather conditions it is likelythat photochemical remineralisation is a significant sourceof nitrogen to phytoplankton and bacteria in surfacewaters. Future research should examine seasonal vari-ability in photoammonification and aim to assess theimportance of this process on marine nitrogen budgets.

Acknowledgements

This work was funded byNordic Council ofMinisters(NMR) grant DONKEY and the Danish ResearchCouncil grant to CS (274-05-0064). We would like tothank Maaret Kulovaara and the Oulu and KouvolaEnvironmental Centres for help in collecting the Finnishsamples, M. Søndergaard for use of the quartz tubes, A.Vähätalo for guidance with apparent quantum yieldcalculations and G. Weidong, R Cory and two reviewersfor their constructive comments on the manuscript.

References

Andersson, C.A., Bro, R., 2000. The N-way toolbox for MATLAB.Chemom. Intell. Lab. Syst. 52, 1–4.

Benner, R., 2002. Chemical composition and reactivity. In: Hansell,D.A., Carlson, C.A. (Eds.), Biogeochemistry of Marine DissolvedOrganic Matter. Academic Press, USA, pp. 59–90.

Bertilsson, S., Stepanauskas, R., Cuadros-Hansson, R., Granéli, W.,Wikner, J., Tranvik, L., 1999. Photochemically induced changes in

bioavailable carbon and nitrogen pools in a boreal watershed.Aquat. Microb. Ecol. 19, 47–56.

Blough, N.V., Del Vecchio, R., 2002. Chromophoric DOM in thecoastal environment. In: Hansell, D.A., Carlson, C.A. (Eds.),Biogeochemistry of Marine Dissolved Organic Matter. AcademicPress, USA, pp. 509–546.

Bro, R., 1997. PARAFAC. Tutorial and applications. Chemom. Intell.Lab. Syst. 38, 149–171.

Bronk, D., 2002. Dynamics of DON. In: Hansell, D.A., Carlson, C.A.(Eds.), Biogeochemistry of Marine Dissolved Organic Matter.Academic Press, USA, pp. 153–249.

Buffam, I., McGlathery, K.J., 2003. Effect of ultraviolet light ondissolved nitrogen transformations in coastal lagoon water.Limnol. Oceanogr. 48, 723–734.

Bushaw, K.L., Zepp, R.G., Tarr, M.A., Schultz-Jander, D.,Bourbonniere, R.A., Hodson, R.E., Miller, W.L., Bronk, D.A.,Moran, M.A., 1996. Photochemical release of biologically avail-able nitrogen from aquatic dissolved organic matter. Nature 381,404–407.

Carstensen, J., Frohn, L.M., Hasager, C.B., Gustafsson, B.G., 2005.Summer algal blooms in a coastal ecosystem: the role of atmo-spheric deposition versus entrainment fluxes. Estuar. Coast. ShelfSci. 62, 595–608.

Coble, P.G., 1996. Characterisation of marine and terrestrial DOM inseawater using excitation-emission matrix spectroscopy. Mar.Chem. 51, 325–346.

Cory, R.M., McKnight, D.M., 2005. Fluorescence spectroscopyreveals ubiquitous presence of oxidized and reduced quinones indissolved organic matter. Environ. Sci. Technol. 39, 8142–8149.

Gao, H., Zepp, R.G., 1998. Factors influencing photoreactions ofdissolved organic matter in a coastal river of the SoutheasternUnited States. Environ. Sci. Technol. 32, 2940–2946.

Grasshoff, K., Voipio, A., 1981. Chemical oceanography. In: Voipio,A. (Ed.), The Baltic Sea. Elsevier Oceanography Series. Elsevier,The Netherlands, pp. 183–218.

Grzybowski, W., 2003. Are data on light-inducted ammonium releasefrom dissolved organic matter consistent? Chemosphere 52,933–936.

Hertel, O., Ambelas Skjøth, C., Brandt, J., Christensen, J.H., Frohn,L.M., Frydendal, J., 2003. Operational mapping of atmosphericnitrogen deposition to the Baltic Sea. Atmos. Chem. Phys. 3,2083–2099.

Holmes, R.M., Aminot, A., Kérouel, R., Hooker, B.A., Peterson, B.J.,1999. A simple and precise method for measuring ammonium inmarine and freshwater ecosystems. Can. J. Fish. Aquat. Sci. 56,1801–1808.

Hu, C., Muller-Karger, F.E., Zepp, R.G., 2002. Absorbance,absorption coefficient, and apparent quantum yield: a commenton common ambiguity in the use of these optical concepts. Limnol.Oceanogr. 47, 1261–1267.

Jørgensen, N.O.G., Tranvik, L., Edling, H., Graneli, W., Lindell, M.,1998. Effects of sunlight on occurrence and bacterial turnover ofspecific carbon and nitrogen compounds in lake water. FEMSMicrobiol. Ecol. 25, 217–227.

Kieber, R.J., Hydro, L.H., Seaton, P.J., 1997. Photooxidation oftriglycerides and fatty acids in seawater: Implication toward theformation of marine humic substances. Limnol. Oceanogr. 42,1454–1462.

Koopmans, D.J., Bronk, D.A., 2002. Photochemical productionof dissolved inorganic nitrogen and primary amines from dissolvedorganic nitrogen in waters of two estuaries and adjacent surficialgroundwaters. Aquat. Microb. Ecol. 26, 295–304.

239C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240

Autho

r's

pers

onal

co

py

Kouassi, A.M., Zika, R.G., 1990. Light-induced alteration ofphotophysical properties of dissolved organic matter in seawater(part 1). Neth. J. Sea Res. 27, 25–32.

Kowalczuk, P., Stedmon, C.A., Markager, S., 2006. Modelingabsorption by CDOM in the Baltic Sea from season, salinity andchlorophyll. Mar. Chem. 101, 1–11.

Lakowicz, J.R., 1983. Principles of Fluorescence Spectroscopy.Plenum, NewYork.

McKnight, D.M., Boyer, E.W., Westerhoff, P.K., Doran, P.T., Kulbe,T., Andersen, D.T., 2001. Spectrofluorometric characterisation ofdissolved organic matter for indication of precursor organicmaterial and aromaticity. Limnol. Oceanogr. 46, 38–48.

Meybeck, M., 1982. Carbon nitrogen and phosphorus transport byworld rivers. Am. J. Sci. 282 (4), 401–450.

Miller, W.L., Zepp, R.G., 1995. Photochemical production of dissolvedinorganic carbon from terrestrial organic matter: significance to theoceanic organic carbon cycle. Geophys. Res. Lett. 22, 417–420.

Mobed, J.J., Hemmingsen, S.L., Autry, J.L., McGown, L.B., 1996.Fluorescence characterisation of IHSS humic substances: totalluminescence spectra with absorbance correction. Environ. Sci.Technol. 30, 3061–3065.

Moran, M.A., Zepp, R.G., 1997. Role of photoreactions in theformation of biologically labile compounds from dissolved organicmatter. Limnol. Oceanogr. 42 (6), 1307–1316.

Murphy, K.R., Ruiz, G.M., Dunsmuir, W.T.M., Waite, T.D., 2006.Optimised parameters for rapid fluorescence-based verification ofballast water exchange by ships. Environ. Sci. Technol. 40,2357–2362.

Peuravuori, J., Pihlaja, K., 1997. Molecular size distribution andspectroscopic properties of aquatic humic substances. Anal. Chim.Acta 337, 133–149.

Rosenfeld, J.K., 1979. Ammonium adsorption in nearshore anoxicsediments. Limnol. Oceanogr. 24, 356–364.

See, J.H. 2003. Availability of humic nitrogen to phytoplankton. PhDdissertation Thesis. College of William and Mary.

Smith, R.C., Baker, K., 1981. Optical properties of the clearest naturalwaters. Appl. Opt. 20, 177–184.

Stålnacke, P., Grimvall, A., Sundblad, K., Tonderski, A., 1999.Estimation of riverine loads of nitrogen and phosphorous to theBaltic Sea, 1970–1993. Environ. Monit. Assess. 58, 173–200.

Stedmon, C.A., Markager, S., 2001. The optics of chromophoricdissolved organic matter (CDOM) in the Greenland Sea: Analgorithm for differentiation between marine and terrestriallyderived organic matter. Limnol. Oceanogr. 46, 2087–2093.

Stedmon, C.A., Markager, S., 2005a. Resolving the variability indissolved organic matter fluorescence in a temperate estuary andits catchment using PARAFAC analysis. Limnol. Oceanogr. 50,686–697.

Stedmon, C.A., Markager, S., 2005b. Tracing the production anddegradation of autochthonous fractions of dissolved organic matterby fluorescence analysis. Limnol. Oceanogr. 50, 1415–1426.

Stedmon, C.A., Markager, S., Bro, R., 2003. Tracing dissolved organicmatter in aquatic environments using a new approach tofluorescence spectroscopy. Mar. Chem. 82, 239–254.

Stedmon, C.A., Markager, S., Søndergaard, M., Vang, T., Laubel, A.,Borch, N.H., Windelin, A., 2006. Dissolved organic matter (DOM)export to a temperate estuary: seasonal variations and implicationsof land use. Estuar. Coast. 29 (3), 388–400.

Stepanauskas, R., Jørgensen, N.O.G., Eigaard, O.R., Zvikas, A.,Tranvik, L., Leonardson, L., 2002. Summer inputs of riverinenutrients to the Baltic Sea: bioavailability and eutrophicationrelevance. Ecol. Monogr. 72, 579–597.

Søndergaard, M., Stedmon, C.A., Borch, N.H., 2003. Fate ofterrigenous dissolved organic matter (DOM) in estuaries: aggre-gation and bioavailability. Ophelia 57, 161–176.

Vähätalo, A.V., Zepp, R.G., 2005. Photochemcial mineralisation ofdissolved organic nitrogen to ammonium in the Baltic Sea.Environ. Sci. Technol. 39, 6985–6992.

Vähätalo, A.V., Salonen, K., Münster, U., Jävinen, M., Wetzel, R.G.,2003. Photochemical transformation of allochthonous organicmatter provides bioavailable nutrients in a humic lake. Arch.Hydrobiol. 156, 287–314.

Wiegner, T.N., Seitzinger, S.P., 2001. Photochemical and microbialdegradation of external dissolved organic matter inputs to rivers.Aquat. Microb. Ecol. 24, 27–40.

240 C.A. Stedmon et al. / Marine Chemistry 104 (2007) 227–240