This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Chromophoric dissolved organic matter (CDOM) variability in Barataria Basin usingexcitation–emission matrix (EEM) fluorescence and parallel factoranalysis (PARAFAC)

Shatrughan Singh a, Eurico J. D'Sa a,⁎, Erick M. Swenson b

a Dept. of Oceanography & Coastal Sciences, Coastal Studies Institute, Louisiana State University, Baton Rouge, LA 70803, USAb Dept. of Oceanography & Coastal Sciences, Louisiana State University, Baton Rouge, LA 70803, USA

a b s t r a c ta r t i c l e i n f o

Article history:Received 10 October 2009Received in revised form 18 March 2010Accepted 25 March 2010Available online 5 May 2010

Keywords:CDOMLouisianaBarataria BasinPARAFACEEM

Chromophoric dissolved organic matter (CDOM) variability in Barataria Basin, Louisiana, USA,was examined byexcitation emissionmatrix (EEM) fluorescence combinedwith parallel factor analysis (PARAFAC). CDOMopticalproperties of absorption andfluorescence at 355 nmalong anaxial transect (36 stations) duringMarch, April, andMay 2008 showed an increasing trend from the marine end member to the upper basin with mean CDOMabsorption of 11.06±5.01, 10.05±4.23, 11.67±6.03 (m−1) and fluorescence 0.80±0.37, 0.78±0.39, 0.75±0.51 (RU), respectively. PARAFAC analysis identified two terrestrial humic-like (component 1 and 2), one non-humic like (component 3), and one soil derivedhumic acid like (component 4) components. The spatial variationof the components showed an increasing trend from station 1 (near the mouth of basin) to station 36 (endmember of bay; upper basin). Deviations from this increasing trend were observed at a bayou channel with veryhigh chlorophyll-a concentrations especially for component 3 in May 2008 that suggested autochthonousproduction of CDOM. The variability of componentswith salinity indicated conservativemixing along themiddlepart of the transect. Component 1 and 4were found to be relatively constant, while components 2 and 3 revealedan inverse relationship for the sampling period. Total organic carbon showed increasing trend for each of thecomponents. An increase in humification and a decrease in fluorescence indices along the transect indicated anincrease in terrestrial derived organicmatter and reducedmicrobial activity from lower to upper basin. The use ofthese indices along with PARAFAC results improved dissolved organic matter characterization in the BaratariaBasin.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Chromophoric dissolved organic matter (CDOM) is the componentof dissolved organic matter (DOM) that absorbs light in the ultravioletand visible range of the electromagnetic spectrum. CDOM present ashumic and fulvic acids, represents a large fraction of dissolved organicmaterial that contributes to light absorption and fluorescence by non-chlorophyllousmaterial in a coastal region (Nieke et al., 1997). Changesin CDOM absorption and fluorescence reflect the variations in CDOMcomposition from several autochthonous or allochthonous sourcesresulting from physical, biological, and chemical processes that occur inthe water column (Coble, 1996; McKnight et al., 2001; Stedmon et al.,2007). Light absorption by CDOM can influence primary productivity inaquatic ecosystems and plays a key role in photo-chemically inducedtransformations in surface waters. Spectral measurements of CDOMabsorption and fluorescence provides useful information of CDOMcharacteristics in estuaries (Stedmon et al., 2006), coastal waters (D'Sa

et al., 2006; D'Sa & Miller, 2003; D'Sa, 2008) and oceans (Coble et al.,1998). Fluorescence spectroscopy technique by means of excitation–emission matrices (EEMs) has been widely used to characterize CDOMin coastal and oceanic environments (Nieke et al. 1997; Coble, 1996;McKnight et al., 2001; Stedmon and Markager, 2005a). However, toovercome the difficulty in identifying individual fluorescent compo-nents in a water sample and to compare the studies related withdissolved organicmatter done in different regimes, statistical tools suchas principal component analysis (PCA) (Boehme et al., 2004) andparallel factor analysis (PARAFAC) have been used (Stedmon et al.,2003; Cory and McKnight, 2005; Kowalczuk et al., 2009).

The EEMs in conjunction with PARAFAC have been used to tracephotochemical and microbial reactions with organic matter (Cory andMcKnight, 2005; Stedmon and Markager, 2005b), in water sourcecategorization (Hua et al., 2007), and correlation with water qualityparameters (Holbrook et al, 2006). It has recently been applied tocharacterize colored dissolved organic matter in lakes (Cory andMcKnight, 2005; Zhi-gang et al., 2007), estuaries (Stedmon et al.,2003; Stedmon et al., 2005a), rivers (Hua et al., 2007), bays (Lucianiet al., 2008; Yamashita et al., 2008) andwetlands (Holbrook et al., 2006;Fellman et al., 2008). PARAFAC modeling provides unique solution to

Science of the Total Environment 408 (2010) 3211–3222

⁎ Corresponding author. Tel.: +1 225 578 0212.E-mail address: ejdsa@lsu.edu (E.J. D'Sa).

0048-9697/$ – see front matter © 2010 Elsevier B.V. All rights reserved.doi:10.1016/j.scitotenv.2010.03.044

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r.com/ locate /sc i totenv

Author's personal copy

dissolved organic matter EEM dataset and is argued as an importantanalytical tool to characterize the CDOM in complex coastal regimes.Two primary fluorescing groups in dissolved organic matter studieshave been identified as humic-like and protein-like substances. Humicsare complexmixture of aromatic and aliphatic compoundsderived fromdecay of organic matter while protein-like substances are associatedwithhighbiological activity.Humic substances are further characterizedbyhumic acids and fulvic acidsmainly differentiatedon thebasis of theirsolubility in alkaline and acidic solutions, respectively (Harvey andBoran, 1985). Humics precipitate when treated with low pH acidicsolutions whereas fulvics remain dissolved in acidic solutions. Thepresence of protein-like substances generally implies autochthonousproduction of CDOM and presence of microbial activity (Stedmon et al.,2007). PARAFACmodeling aids to efficiently resolve these two classes offluorophores and identify the dominance of a particular class in anenvironment.

The Barataria Basin is a large complex, well-mixed, estuarine systemlocated south of New Orleans, Louisiana, USA and directly west of themouthof theMississippi River. The basin is influencedby theMississippiRiver through exchange of river water at the marine end in the lowerbasin and also through various siphon points associated with diversionof river water into the basin to accommodate flood control, navigationand coastal restoration. Swamp forests and marshes characterize thenature of the Barataria Basinwhich is isolated from theMississippi riverdue to levee construction since the 1930s (Mossa, 1996), laterreconnected by Davis pond diversion. Various studies have addressedthe impact of the diversion of the Mississippi River into the BaratariaBasin on the habitat, organic matter, and water quality of the Basin(Swenson et al., 2006; Wissel et al., 2005; Jones et al., 2002; Conner &Day, 1987). Marsh derived soil organic matter from Barataria Basin hasbeen shown to contribute to the total organic carbon pool throughresuspension of unconsolidated sediments and then exported towardsadjacent Louisiana shelf waters (Sampere et al., 2008). Studies have

been conducted on carbon and nitrogen flow from the Barataria Basininto the Gulf of Mexico (Feijtel et al., 1985; Castro et al., 2003) withmodeling of the Basin showing it to be a small source of total organiccarbon to the northern Gulf ofMexico (Das et al., 2009). The objective ofthe present study is to characterize the fluorescent portion of CDOM inBarataria Basin using field data acquired along an axial transect duringthree field campaigns in the spring of 2008 that corresponded to thehigh flow discharge from the Mississippi River and its controlleddiversion into the basin especially during high river stages (Fig. 1). EEMand PARAFAC analysis are used to examine compositional distributionand CDOM variation with PARAFAC sample scores being used toexamine probable linkages to wetlands, agricultural sources and otherwater bodies.

2. Methods

2.1. Sampling site

The Barataria Basin is located west of theMississippi River delta andis separated from the Gulf of Mexico in the south by the Grande TerreIslands and by the Mississippi River in the north. It is a systemmade upof biologically rich and productive habitats including swamps, fresh,brackish, and salinemarshes, bayous, bays, and barrier islands. BaratariaBasin is an irregular shaped basin covering 1673 km2 surface area and5700 km2 of drainage area with an average depth of 2.0 m (Fig. 1). Theaverage daily fresh water inflow into the basin is approximately156 m3 s−1 with an average salinity of 13. The main freshwater sourcesfor the Barataria Basin include rainfall, local run-off, Intracoastalwaterway that delivers water from Atchafalaya River during high flowconditions to Mississippi River, and diversions from Davis Pond with adesign-pumping rate of 300 m3 s−1, Naomi andWest Pointe A la Hachewith a maximum pumping rate of 60 m3 s−1 at each site, and PortSulphur. Apart from these freshwater sources, numerous major and

Fig. 1. Location of surface water sampling stations in Barataria Basin, Louisiana, USA along a 124 km long axial transect. Filled circles represent the stations with large size circleshaving annotations. Distances between annotated stations are shown.

3212 S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

minor point discharges, septic tanks, sewage/stormwater overflow,unsewered communities, pasture lands, and marshes contribute tochanges in water quality of the basin (Conner & Day, 1987). Althoughthe salinity signals are highly coherentwithMississippi River discharge,these point sources can significantly contribute to a decrease in theaverage salinity of the basin (Swenson et al., 2006, Inoue et al., 2008).

2.2. Collection of water samples

Surface water samples were collected at 36 stations (Fig. 1) locatedalong a 124 km transect (South-East to North-West) from the mouthof the basin (station 1) in the Gulf of Mexico to the north-west end ofthe upper basin (station 36) duringmonthly field trips inMarch, April,and May, 2008. Salinity profiles were recorded using a handheld YSI.The surface water samples collected at 36 sampling stations werebrought back to the laboratory and immediately filtered using pre-rinsed 0.2 micron nucleopore membrane filters. The filtered sampleswere kept in a refrigerator for optical absorption and fluorescenceanalyses. The tap water was treated by a Barnstead Nanopure®ModelD-50280 purification system with purity of 18.2 MΩ Milli-Q. Mea-surements of CDOM absorption and fluorescence were made in thelaboratory. Humification Index (HIX) (Zsolnay et al., 1999) wascalculated by dividing the integrated area of fluorescence intensitymeasured between emission wavelength of 435 and 480 nm by theintegrated area of fluorescence intensity measured between emissionwavelength of 300 and 345 nm with excitation at 254 nm. Fluores-cence Index (FI) (Cory and McKnight 2005), calculated as a ratio ofmaximum emission fluorescence intensities at 470 and 520 nm,excited at 370 nm (e.g., Tzortziou et al., 2008) was used to distinguishsources of dissolved organicmatter containing humic substances in anaquatic environment with suggested range for terrestrially originatedhumics of 1.4 and for marine originated materials of 1.9.

2.3. Absorption spectroscopy

The surface water samples were allowed to reach room temper-ature and the instrument was kept on for 30 min before the sampleanalysis. Absorption spectra were obtained between 190 and 750 nmat 2-nm intervals using Perkin Elmer Lambda 850 double-beamspectrophotometer equipped with 1 cm path-length quartz cuvette(volume of 4 ml) and 150 mm spectralon coated integrating sphere.The data were corrected for scattering and baseline fluctuations bysubtracting the average value of absorption between 700 and 750 nmfrom each spectrum. The absorption coefficients (a) were calculatedfrom the absorbances (A) obtained from the spectrophotometerusing:

aðλÞ = 2:303 ×AðλÞl

where A(λ) is the absorbance at a wavelength λ, and l is the path-length in meters.

2.4. Fluorescence spectroscopy

Surface water samples were treated in a similar manner as those forabsorption measurements. Samples having absorbance greater than0.02 at 350 nm (A350N0.02 for 1 cm pathlength) were diluted withparticle freeNanopureMilli-Qwater (also used asablank) to account forthe inner filter effects. However, any inner filter effects at wavelengthsless than 350 nm were assumed to be negligible. Excitation Emissionmatrices (EEM)were generated using a Horiba Jobin Yvon Fluoromax-4spectrofluorometer equipped with a 50W ozone-free Xenon arc lampand a R928P photomultiplier tube as a detector. The spectrofluorometerwas set to collect the signal in ratiomodewith dark offsets using a 5 nmbandpass on the excitation as well as emission monochromators.

Factory supplied correction factors were applied to the scans to correctfor instrument configuration. The EEMswere recorded at 5 nm intervalsfor excitation spectra between 250 and 500 nm (Zepp et al., 2004;Kowalczuk et al., 2005) and emission spectra between 280 and 600 nmat an integration time of 0.1 s. FLToolbox (ver. 2.10b; February, 2007)developed by Wade Sheldon (University of Georgia) for MATLAB®(Zepp et al., 2004) was utilized for post processing of data. In brief, thesoftware removes the Rayleigh and Raman scattering peaks by excisingportions (± 10–15 nm FW) of each scan centered on the respectivescatter peak and then replaces the excised data with interpolated datausing the Delaunay triangulation method. Subsequently, Milli-Q waterblank EEMswere subtracted from the sample EEMs to eliminate Ramanpeaks and then the EEMs were normalized to daily-determined waterRaman integrated area maximum fluorescence intensity (350 ex/397em, 5 nm bandpass) (Colin Stedmon, pers. comm.). Finally, the EEMswere multiplied with a dilution factor derived from the fluorescenceintensity at Ex/Em=350/397 nm of water Raman scan to obtain theintensity for the original, undiluted sample (Coble et al., 1998). Thefluorescence intensities measured were reported in Raman Units (RU)in this study. A 5% agreement was noted between replicate scans interms of intensity and within bandpass resolution in terms of peaklocation using Milli-Q water scans. Four samples for May (stations 31,32, 34, and 35) could not be run because of unavailability of theinstrument.

2.5. Chlorophyll-a and Total Organic Carbon (TOC) measurements

Phytoplankton chlorophyll-a (Chl-a) containing planktons weremeasured using EPA method 445.0 (Arar and Collins, 1992). Theplanktons were concentrated from a volume of water by filtering at alow vacuum through a glass fiber filter (GF/F). The pigments wereextracted from the phytoplankton using a solution of 90% Acetone and10% dimethyl sulfoxide (DMSO). The samples were allowed to steepfor 2 to 24 h (maximum) to extract the Chl-a. The samples were thencentrifuged to clarify the solution. The fluorescence values weremeasured before and after acidification with 0.1 N HCl. The fluores-cence readings were then used to calculate the Chl-a concentrations(mg m−3) in the sample extracts.

Total carbon (TC) was measured by employing High TemperatureCatalytic Oxidation (HTCO) using a Shimadzu® TOC-5000A Analyzer.The machine operates by combusting the water sample (at 680°centigrade) in a combustion tube filled with a platinum-aluminacatalyst. The carbon in the sample is combusted to become CO2 whichis detected by a non-dispersive infrared gas analyzer (NDIR) to givethe total amount of carbon in the sample. Inorganic carbon (IC) isanalyzed by first treating the sample with Phosphoric acid to obtainthe total amount of inorganic carbon in the sample. Total OrganicCarbon (TOC) is obtained by subtracting the IC value from the TCvalue.

2.6. The PARAFAC model and analysis

PARAFAC statistically decomposes the three-way data intoindividual fluorescence components. Bro (1997) has demonstratedthe uniqueness of PARAFAC decomposition method to reduce an EEMinto trilinear terms and a residual array using:

Xijk = ∑F

n=1ainbjnckn + εijk

where, for EEMfluoroscence data,Xijk is thefluorescence intensity of theith sample at the kth excitation and jth emission wavelength. ain isdirectly proportional to the concentration (e.g., moles C; here defined asscores) of the nth fluorophore in the ith sample. bjn and ckn are estimatesof emission and excitation spectra (loadings) of nth fluorophore atwavelength j and k, respectively. F is the number of components

3213S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

(fluorophores) and εijk the residual matrix of the model whichrepresents unexplained variability by the model. Holbrook et al.(2006) suggested that individual components from a large complexEEM dataset can be identified provided the correct number ofcomponentswere chosen after outlierswere removedwith preliminaryexamination of dataset. The PARAFAC model identifies the number ofcomponents as well quantifies the scores for each component that isdirectly proportional to the the component's (fluorophore) concentra-tion in the sample. This concentration can be converted into actualconcentration provided specific absorption coefficient or quantumyieldassociated with excitation and emission of each fluorophore is known.

PARAFAC analysis was performed in MATLAB ver., 7a using theDOMFluor toolbox (ver., 1.7; Feb., 2009) developed for MATLAB® byColin Stedmon (NERI, Aarhus University, Denmark). PARAFAC con-straints, such as nonnegativity, and model initialization values derivedfrom singular value decomposition (SVD) were used following themethod of Stedmon et al. (2003, 2008). PARAFAC was applied to ourEEM dataset (104 samples×65 emissions×51 excitations) with someanalytical and statistical assumptions as reported by Stedmon and Bro(2008). Determination of the number of components (i.e., modelvalidation) was done by split half analysis and analysis of residuals andloadings (Stedmon et al., 2003). The number of components were alsoverified by CORCONDIA (Core Consistency Diagnostic) and JackknifingTechniques (Riu and Bro, 2003) that show similar explained variability.Based on the relative decrease in core consistency, four componentswere identified for the dataset with some variability remaining in theresiduals. Afive-componentPARAFACmodelwas rejectedas it couldnotbe validated using split-half analysis. Further, close similarity in thePARAFAC excitation-emission loadings obtained for a single split-halfversus the whole dataset (Fig. 5) suggest that the selection of the 4-component model in this study was justified. A typical example of themeasured, modeled and residual EEM data is given in Fig. 2. It appearsfrom Fig. 2 that the four-component PARAFACmodel captured the bulkfeatures in the measured EEM as indicated by the low residual(difference between measured and modeled data). However, somesmall peaks seennearfirst order Rayleigh scattering trough could not be

modeled. This could be attributed to the size of the dataset or to theproximity of missing value settings done prior to running the PARAFACmodel. The missing values lie near the Raman and Rayleigh scatteringand were cut to avoid any influence by these values in the final dataset.The small peaks in Fig. 2 may also indicate the presence of weakfluorescing components due to noise, trivial quenching effects, innerfilter or scattering effects. Fmax was identified for each water samplecollected over the Barataria Basin transect for the three months —

March, April, and May, 2008. Table 1 shows the mean Fmax valuesdetermined over the threemonth period for each component identifiedby PARAFAC model in this study using the peak (maximum) emissionand excitation wavelengths. We have used leverage and loadingtechnique (Stedmon and Bro, 2008) to identify the outliers and verifiedthem using Jack-Knife technique (Riu & Bro, 2003). Seven samples(March: stations 5, 31, 33; April: station 31; May: stations 19, 33, and36) and three emission wavelengths (280, 285, and 290 nm) wereremoved from thewhole EEMdataset before running the final PARAFACmodel on afinal dataset (97 samples×62 emissions×51excitations) forfour individual components.

3. Results and discussion

3.1. CDOM absorption and fluorescence variability along the transect

The monthly variation of CDOM absorption and fluorescence at355 nm along the Barataria Basin transect (3a, b) and its variation withsalinity (Fig. 3c) show similar increasing trends for March, April, andMay from themouth of the basin (station 1-marine endmember) to theend of the transect at upper basin (station 36-terrestrial end member)with some exceptions in May. On average CDOM absorption (fluores-cence) increased from a low of 3.25 m−1 (0.20 RU) at st. 1 (salinity16.07) to a high of 22.22 m−1 (1.57 RU) at st. 35 (salinity 0.11).Exceptions inMaywere observedbetween station 24 to 30,where pointdischarges from nearby living communities and runoff from freshmarshes into the narrow bayou channel could be responsible forobserved variability in this part of the transect (discussed later). An

Fig. 2. Typical example of surface and contour plot of a measured EEM and PARAFACmodeling result for station 15 (amiddle basin station) showingmeasured, modeled, and residualEEM. Note the change in z-axis representing fluorescence reported in Raman Units (RU).

3214 S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

unexplained variability of CDOM was observed in the absorption datafor March between stations 34 and 36 where CDOM fluorescenceappeared to follow the increasing trend fromstations34and36whereasa negative trend is observed for the absorption data. Mean CDOMabsorption and fluorescence at 355 nmwere 11.06±5.01, 10.05±4.23,11.67±6.03 (m−1) and 0.80±0.37, 0.78±0.39, 0.75±0.51 (RU) forMarch, April, and May, respectively.

CDOM absorption and fluorescence showed a similar negativerelation with salinity along the transect (Fig. 3c). A conservativerelationship of CDOM optical properties (absorption and fluorescence)with salinity occured in the middle part of the transect in the salinityrange 11.4 to 0.7 and 6.7 to 3.5 for March and April, and 9.8 to 1.2 forMay, respectively. Few exceptions from the general trend can be seen inApril at stationswith high salinity and lowCDOM(Fig. 3c). The intrusionof high salinity sea water into the basin area due to tidal or wind effectsprobably caused a change in salinity dynamics thereby affecting theCDOMvariability in this part of the transect forMarch andMaywhereasin April, lower salinity in this part of the transect could be attributed tofreshwater discharges from upper part of the transect.

3.2. Component variability along the transect

Although four individual components were determined for thisdataset using the PARAFAC model (Fig. 4), it does not suggest thatonly four types of fluorophores were present in these samples. It alsodoes not suggest that all the four components are present in eachsample of the transect. However, it does suggests that these fourcomponents were present in majority of the transect samples. Fig. 5explains the excitation and emission loadings identified for the fourcomponent PARAFAC model possessing excitation spectra with twomaxima and emission spectra with single maxima. A small peakobserved in emission spectra for component 2 can be attributed toartifacts (Fig. 5b). We attempted to remove this by increasing themissing value region (Stedmon, pers. comm.), but with no success.Component 1 has a primary (and secondary) fluorescence peak at anexcitation/emission wavelength of 250 (340) nm/440 nm. Theprimary and secondary fluorescence peaks for component 2 werered shifted compared to component 1, occuring at 270 (370) nm/460 nm. The blue shift for component 3 with respect to component 1and component 2, flourescing at a primary (secondary) excitation/emission wavelength pair of 250 (285) nm/395 nm, is also observed.This component is similar to a non-humic and lies in betweentryptophan and N-peak (Table 2). Component 4 possessed a primary(secondary) excitation/emission wavelength pair of 240 (410) nm/520 nm. This component likely derived from agricultural catchmentsas reported in other estuarine environments (e.g., Stedmon andMarkager, 2005a), and is known to be rapidly produced and removedmaking it difficult to identify its presence. Also, the close proximity ofthe component 4 peak with the terrestrially derived organic materialmakes its identification difficult (Stedmon and Markager, 2005b).

Components 1 and 2 are similar to humic-like peaks as reported inprevious literature (Coble, 1996; Stedmon et al., 2003; Stedmon et al.,2007). Barataria Basin is surrounded by swamps, marshes, forested andwetland regions and the acquired EEMs reflect their influence.Component 1 thought to be A-peak has maximum emission at440 nm, whereas component 2 is found to be similar to previouslyreported C-peak using traditional “peak picking” technique withmaximum emission intensity at 460 nm (Figs. 4 and 5; Tables 1 and2) (Coble, 1996). Component 3 reported as N-peak (Coble et al., 1998),could be a combination of tryptophan (T-peak) and biologically labilematter produced in the surface water. Recently, it is reported as a non-humic component from freshly produced biologically labile material(e.g., Yamashita et al., 2008) and protein or amino-acid like component(Hua et al., 2007; Kowalczuk et al., 2009) (Figs. 4 and 5; Table 2).Component 4 is similar to soil derived humic acid originating fromwetlands and marshes with a primary (secondary) excitation wave-length at 250 (410) and emission wavelength at 520 nm. Thiscomponent is very similar to that found by Lochmuller and Saavedra(1986) in a study done for soil derived fulvic acid formally known as“Contech” FA (ex/em=390/509 nm), but with a red shift in excitationand emissionmaximawhich suggests high aromatic content and higherdegree of humification in our study (Figs. 4 and 5; Tables 1 and 2). Thiscomponent is common to several fresh water regimes and microbial

Table 1Positions of fluorescence maxima and fluorescence intensities, Fmax, (scores) in RamanUnits (R.U.) of the four components identified by PARAFAC analysis in this study.

Component Excitation maxima (nm) Emission maxima (nm) Fmax (R.U.)

Component 1 b250 440 1.16±0.65Component 2 270 (370) 460 0. 78±0.41Component 3 b250 (285) 395 0. 78±0.40Component 4 b250 (410) 520 0. 40±0.18

Fig. 3. Spatial distribution of (a) CDOM absorption coefficient at 355 nm, and (b) CDOMfluorescence at 355 nm. (c) Variation of CDOM absorption and fluorescence with salinity,along the axial transect for March, April, and May, 2008. Empty circles, triangles, andsquares represent CDOM absorption for March, April, and May, respectively, while filledcircles, triangles, and squares represent CDOM fluorescence for similar months.

3215S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

degradation could be a source of this component when exposed tovisible light (Stedmon and Markager, 2005b). Spectral signature of thesoil derived humic acid component (component 4) could also beattributed to the leaching processes from agricultural catchments giventhe long residence time of water in these catchments (Stedmon andMarkager, 2005a).

3.3. Spatio-temporal variability of components

The spatial distribution of components 1, 2, 3, and 4 for the threesamplingmonths (Fig. 6) show similar trends as FCDOM(355 nm) alongthe transect. Component 3 also shows similar trend for most part ofthe transect, except between stations 24 to 30 for May (Fig. 6,

Fig. 4. The four different fluorescent components found by the PARAFAC model. Positions of their maxima are given in Table 2.

Fig. 5. Excitation and emission loadings derived from the four component PARAFAC model using split half validation technique. Solid red lines represent emission loadings for thewhole dataset and dotted red lines represent excitation loadings for thewhole dataset. Dashed blue lines shows emission loadings and dotted blue lines shows excitation loadings fora single split-half from the whole dataset.

3216 S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

enclosed rectangle). Fig. 7 shows the plots of components against eachother for the three month data set that reveals the variability in thefluorescence components. Components 1 and 2 were almost perfectlylinearly correlated (r2=0.99) followed by components 2 and 4(r2=0.97) indicating these components to have a common factorcontrolling their concentrations (Stedmon and Markager, 2005a).However, while components 3 and 1 and components 3 and 2 werealsowell correlated (r2=0.95 and 0.96, respectively), a greater scatteris observed in the data that is enhanced between stations 24 to 30(Fig. 7). This difference in behaviour of specific component from thegeneral trend could be achieved by the use of PARAFAC techniquewhich helped identify difference in component 3 that was notapparent in the single fluorescence emission (Fig. 3c). The presence ofcomponent 3 at these locations is complemented by the increasedChl-a and TOC values (Figs. 8 and 9) suggesting the production ofbiologically labile organic material. The biological production of labileorganic matter in this part of the transect could be associated withperiphytons or macrophytes found in shallow aquatic systems(Kortelainen et al., 2004).

The temporal variation of four components identified usingPARAFAC model for the three months (March, April, and May) indicatecomponents 1, 2 and 4 to be constant for the threemonth period;meanconcentration scores for component 1 was 16±0.09, 1.17±0.10, and1.15±0.15, for component 2, 0.78±0.06, 0.78±0.07, and 0.76±0.15and for component 4, 0.42±0.03, 0.40±0.03, and 0.38±0.04 for thethe months of March, April, and May, respectively. Mean concentrationfor component 3 in contrast showed an increasing trend being 0.71±0.05, 0.78±0.06, and 0.84±0.10, respectively and appeared to beinfluencedbyChl-a concentrations. The enhancedbiological activity andpresence of component 3 at these locations (Fig. 6), could be due to thelonger residence time of water in the narrow Lac des Allemands bayou(Fig. 1), or nutrient enrichment from agricultural catchments in nearbycommunities. The presence of component 1 and 4 during the threemonth period is significant at the same level and could be associated

Table 2Descriptions of the four components identified by PARAFAC analysis of excitationemission matrices (EEM) fluorescence data in this study and their comparison withpreviously identified components.

Component(this study)

Ex/Em(nm)

Description and references

Component 1 b250(340)/440

Humic-like; Terrestrial originA-peak (Coble, 1996)Component 1 (Stedmon et al., 2003)Component 1 (Cory and McKnight, 2005)Component 4 (Stedmon & Markager, 2005a)Component 5 (Stedmon & Markager, 2005b)Component 1 (Holbrook et al., 2006)Component 3 (Ohno and Bro, 2006)Component 1 (Hua et al., 2007)Component 1 (Stedmon et al., 2007)

Component 2 270(370)/460

Humic-like; Terrestrial origin; related to terrestrialparticulate organic matterC-peak (Coble, 1996)Component 3 (Stedmon et al., 2003)Component 1 (Stedmon & Markager, 2005b)Component 1(Ohno and Bro, 2006)Component 2 (Hua et al., 2007)Component 3 (Yamashita et al., 2008)

Component 3 b250(285)/395

Non-Humic-like; Labile matter; Biologicalproduction in the water columnN-peak (Coble et al., 1998)Component 5(Stedmon et al., 2003)Component 3 (Stedmon & Markager, 2005b)Component 2 (Holbrook et al., 2006)Component 4(Hua et al., 2007)Component 5 (Yamashita et al., 2008)

Component 4 b250(410)/520

Soil fulvic-like; derived from agriculturalcatchments and exists in fresh water environmentsSoil-fulvic-acid (Lochmuller and Saavedra, 1986)Component 2 (Stedmon & Markager, 2005a)Component 7 (Stedmon & Markager, 2005b)Component 2 (Stedmon et al., 2007)

Fig. 6. Spatial distributions of scores obtained from PARAFAC modeling for each component (a) component 1 (C-1), (b) component 2 (C-2), (c) component 3 (C-3), and (d)component 4 (C-4), over the three month period (March, April, and May, 2008). Filled circles and squares represent March and May, respectively while empty triangles representApril. A rectangular box is drawn to show the variability observed in component 3 for May, 2008 data.

3217S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

with the presence of allocthonous dissolved organic matter derivedfrom terrestrial sources and leaching fromsoil derivedhumic acidwhichiswidespread in streams, lakes, bayous andwetlands (e.g., Stedmon andMarkager, 2005b; Holbrook et al., 2006; Kowalczuk et al., 2009).

3.4. Component variability with salinity, Chl-a, and TOC

The behavior of the four components identified in this study wasexamined with changes in salinity gradient (Fig. 8). All the fourcomponents show similar trend as observed for aCDOM (355) andFCDOM (355)with salinity (Fig. 3c). The fluorescence intensity of all thefour components decreased along the salinity gradient in the middlepart of the transect and remained non-conservative in the remaining

part of the transect. However, the four components decreased tolowest values in April (from stations 6 to 12) with a subsequentchange in salinity likely due to the intrusion of high salinity sea waterin this part of the transect. Components 1, 2, 3, and 4 showed a similartrend of higher values at low salinity range (stations 18 to 36) and agradual increase in fluorescence intensities along the salinity gradientfrom marine end member to terrestrial end member (station 17). Nosignificant change in fluorescence intensities were observed for firstfour stations (1, 2, 3, and 4) for all the components. A markeddifference in fluorescence intensity of component 3 at stations 25 to30 in comparison to components 1, 2, and 4 for April andMay (Fig. 8c)could indicate the presence of freshly produced organic matter atthese locations during May (Coble et al., 1998; Stedmon et al., 2003;Yamashita et al., 2008).

The variability of the four fluorescent components with Chl-a andTOC (Figs. 9 and 10) andwith station number and salinity (Figs. 6 and 8)are used examine their contribution to DOM fluorescence. The averagetransect Chl-a was 15.65±8.57 (March), 17.11±9.89 (April), and26.74±26.41 mgm−3 (May), respectively with highest values andvariability in May. While the mean fluorescence components 1, 2 and 4remained constant over the threemonth period, themean component 3increased over the same period suggesting its increase to be related tothe increase in Chl-a. The increased Chl-a supports the case ofautochthonous production of biologically labile matter, mostly prom-inent between stations 25 to 30 likely due to runoff from agriculturalcatchments nearby. Three fold increase from station 24 (8.00 mgm−3)to 25 (24.96 mgm−3) was observed for Chl-a in May, that increased to90.15 mgm−3 at station 30, an almost a ten-fold increase in comparisonto station 24. This implies algal bloom like condition at these locationslikely due to proximity of these stations to cultivated crops andassociated runoffs containing high nutrients. The increased Chl-aassociated with phytoplankton blooms are a potential source ofautochthonous DOM at these locations (Stedmon et al., 2006). TheChl-a variability for March and April were not significant at theselocations. Mean TOC values decreased from March (10.98±4.60) to

Fig. 7. Relationships between the fluorescence of components: C3 versus C1 (solidcircles), C3 versus C2 (empty circles), C1 versus C2 (solid triangles), and C2 versus C4(empty triangles).

Fig. 8. Concentration scores (FMAX) of each PARAFAC component (a) component 1 (C-1), (b) component 2 (C-2), (c) component 3 (C-3), and (d) component 4 (C-4) plotted versussalinity. Concentration scores, Fmax are in Raman Units (RU. Breaks in plots are shown to visualize the variability at low salinity values.

3218 S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

April (8.61±3.68) and later increased in May (9.37±6.44). In general,TOC values increased with increasing component concentration scoresfor March, April, and May (Fig. 10). However, a deviation from thegeneral trend in component 3 with high scores values in comparison tothe other componentswas observed between stations 24 to 30 that alsocontained high TOC concentrations (Fig. 10c). Although measured TOC

values have not shown as large a variation as Chl-a during May atstations 25 to 30, it however remained high at these locationssupporting the argument of autochthonous production of biologicalmaterial (Figs. 10c and 6). Components 1, 2, and 4 have shown similartrendwith Chl-a variation forMay at stations 25 to 30, but fluorescenceintensity of component 3 was higher indicating the biological matter

Fig. 9. Concentration scores (FMAX) of each PARAFAC component (a) component 1 (C-1), (b) component 2 (C-2), (c) component 3 (C-3), and (d) component 4 (C-4) versusChlorophyll-a (Chl-a). Concentration scores, Fmax are in Raman Units (RU), and Chl-a values are in mg m−3.

Fig. 10. Concentration scores (FMAX) of each PARAFAC component (a) component 1 (C-1), (b) component 2 (C-2), (c) component 3 (C-3), and (d) component 4 (C-4) versus totalorganic carbon (TOC). Concentration scores, Fmax are in Raman Units (RU), and TOC values are in mg l−1.

3219S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

production (Fig. 9c). Similarly, a significant increase in concentrations oftotal organic carbon can be seen in May for component 3 at theselocations (stations 25 to 30) in comparison to other components(Fig. 10c).

3.5. Humification and Fluorescence Index

The humification and fluorescence indices have been calculated asdescribed in methods section and plotted in Fig. 11. The humificationindex (HIX), an indicator of dissolved organic matter bioavailabilitywithin a natural system, is used to investigate the degree ofhumification because highly humified organic material is expected

to be less labile in comparison to that of low degree humified organicsubstance (Zsolnay et al., 1999; Ohno, 2002). This also indicate thehigher degree humified organic matter should persist longer thanlower degree humified substance in an environment.

Fluorescence Index (FI), an indicator of dissolved organic matterderived from terrestrial or microbial activity in an environment, isused to measure the difference between fluorescence emissionintensity at two different emission wavelength with a singleexcitation wavelength which then provide information about blueshift or red shift in emission maxima depending on its origin(McKnight et al., 2001). We have observed a value near 1.4 in mostof the monthly samples, indicating mainly terrestrial origin of humicsubstances in March, April, and May. Few values higher than 1.4observed in May at stations close to the marine end member could bedue to microbial activity at these stations (stations 1, 2, 3, and 4).

A negative relationship is obtained for humification and fluores-cence index for most part of the transect gradient (Fig. 11a) and withsalinity (Fig. 11b). In general, humification increased along thetransect as terrestrial derived organic matter increased from lowerto upper basin (from station 1 to station 36). However, a decrease influorescence index is noticed as a general trend from station 1 to 36which implies reducedmicrobial activity along the transect. However,exceptions to these general trends can clearly be seen betweenstations 25 to 30 as a decrease in humification index and an increasein fluorescence index, particularly for May, along the stations(Fig. 11a). Also, an increase in fluorescence index and a decrease inhumification index with salinity gradient (Fig. 11b) can be clearlyseen for first four stations (station 1 to 4). The observed statisticalrelationship obtained for HIX and FI with salinity are [HIX=12.67 –

0.22 * Salinity; p=b0.0001; R=0.5] and [FI=1.36+0.003 * Salinity;p=b0.0001; R=0.6]. Both these non-linear regressions wereperformed at 95% confidence interval.

We have calculated the total normalized fluorescence intensity assuggestedbyKowalczuket al. (2009)by summingfluorescence intensitiesof all the four identified components in each month and plotted themagainst HIX and FI in Fig. 11c. The variation for HIX was interesting toobserve showing a decrease in HIX values from station 25 onwards.However, FI did not show significant increase (neither a decrease) withtotal normalized fluorescence intensity in this part of the transect. Adecrease shown by HIX as a function of total normalized fluorescenceintensity near station 25 onwards again suggested the role of in situphytoplankton production at these locations. The monthly variability inmean values of total normalized fluorescence intensity was observed as3.08±1.26 (March), 3.14±1.83 (April), and 3.13±0.02 (May) in RamanUnits (RU). Similarly,mean values of HIX and FIwere obtained as 13.33±2.14 (March), 11.98±1.83 (April), and 10.25±1.44 (May) for HIX, and1.36±0.02 (March), 1.37±0.03 (April), and 1.38±0.03 (May) for FI,respectively.HIXvalueswereobserveddecreasingwhereas FI valueswerefound increasing over the monthly mean suggesting increased microbialactivity and decreased humification over the study period. It could alsoimply an increase in autochthonous production later in spring. Long termmonitoring would improve an understanding of the seasonal variability.Student's t-test (two-tailed, type 2) were performed (MATLAB® at asignificance level, p-value=0.05) to assess whether HIX and FI valuesdetermined for this study were found to be significantly correlated withmonths. p-values obtained from these analyses showed significantdifferences in HIX calculated for March and April (p-value=0.0025),and forApril andMay(p-value=0.0016). Theobserved level of significantdifference for March and May sampling period yielded p-valueb0.001.These results indicated that the HIX values were significantly differentover the sampling periods. Similarly, FI values were analyzed and onlyMarch and May (p-value=0.002) were found as significantly different.The others (March and April; April andMay) have shown that therewereno significant differences in FI values over the sampling period as theobserved p-values were very high (N0.05). These results suggest that HIXwas more significantly different than FI over the three month sampling

Fig. 11. Humification Indices (HIX) and Fluorescence Indices (FI) versus (a) transectstations, (b) salinity, and (c) the total normalized fluorescence intensity (calculatedfrom summation of concentration scores, FMAX identified for PARAFAC components),Raman Units (RU).

3220 S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

period and explains themore pronounced variability of HIX than FI alongthe transect.

4. Conclusions

Fluorescence spectroscopy applied with PARAFAC technique tostudy colored dissolved organic matter dynamics in Barataria Basinpresented unique results in terms of the behavior of component 3which was not apparent in single emission measurement at 355 nm,and the identification of soil humic component. Further this studyenabled us to quantify the compositional contributions by differentfluorophores in the study area. The fluorescence intensities obtainedfor the identified components showed their distribution along thetransect revealing contribution by autochthonous dissolved organicmaterial in a section of the transect. Results of this study suggest theneed to consider the role of nutrients being discharged into the baythat could contribute to algal bloom formation at various locationsalong the transect. Conservative and non-conservative mixing havebeen observed in different parts of the transect suggesting theinfluence of point discharges. As expected we observed major humic-like fluorescence intensities while in situ production of dissolvedorganic matter in a small part of the transect was reflected in thefluorescent components as well as HIX and FI. In all, four componentsusing PARAFAC model were indentified: two humic-like, one non-humic like, and one soil-humic like. The dominance of humic-likefluorophores in an estuarine region has already been established, butstudies dealing with non-humic components and soil derived humicacids are limited. We not only identified the number of components,but presented their relationships with salinity, Chl-a, and total organiccarbon measured along the transect. HIX and FI, both complementeach other with an inverse relationship and can serve in addition toPARAFAC technique as identification tools for CDOM source and fatein estuarine environments. However, HIX showed a greater range ofvariability than FI for this study which could be considered as a moreuseful index to study CDOM dynamics, especially in coastal andestuarine regions like Barataria Basin. This study supported the use ofexcitation emission matrix fluorescence in combination with parallelfactor analysis to distinguish between different sources of dissolvedorganic matter. Its applicability in estuarine regions such as BaratariaBasin could serve as a useful tool to evaluate the dynamics of dissolvedorganic matter in similar complex coastal regimes.

Acknowledgements

The authors are grateful to Dr. E. Turner from the Department ofOceanography and Coastal Sciences, Louisiana State University for thewater samples, salinity, Chl-a and total organic carbon data. Theauthors are thankful to Dr. Colin Stedmon (NERI, Aarhus University,Denmark) for the help with PARAFAC analysis, interpretation and hisinsightful comments over this manuscript. The authors acknowledgesupport provided by the Minerals Management Service (MMS) —

Coastal Marine Institute (CMI) Cooperative Agreement (1435-0104CA32806), by a NASA grant (NNA07CN12A) to E. D'Sa and by aNorthern Gulf of Mexico Initiative (NGI) — Shell grant.

References

Arar EJ, Collins GB. Method 445.0 In vitro determination of Chlorophyll-a in marine andfreshwater phytoplankton by fluorescence. Environmental Monitoring SystemsLaboratory, Office of Research and Development. Cincinnati, OH: U.S. Environmen-tal Protection Agency; 1992. p. 45268. Reprinted by Turner Designs, Sunnyvale, CA94086.

Boehme J, Coble PG, ConmyR, Stovall-LeonardA. Examining CDOMfluorescence variabilityusing principal component analysis: seasonal and regional modeling of three-dimensional fluorescence in the Gulf of Mexico. Marine Chemistry 2004;89:3-14.

Bro R. PARAFAC. Tutorial and applications. Chemometrics and Intelligent LaboratorySystems 1997;38(2):149–71.

Castro MS, Driscoll CT, Jordan TE, Reay WG, Boynton WR. Sources of nitrogen toestuaries in the United States. Estuaries 2003;26(3):803–14.

Coble PG. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Marine Chemistry 1996;51(4):325–46.

Coble PG, Del Castillo CE, Avril B. Distribution and optical properties of CDOM in theArabian Sea during the 1995 Southwest Monsoon. Deep-Sea Research II 1998;45:2195–223.

Conner WH, Day Jr JW. The ecology of Barataria Basin, Louisiana: an estuarine profile.Fish and Wildlife Service, U.S. Department of the Interior. Biological Report; 1987.85 (7.13).

Cory RM, McKnight DM. Fluorescence spectroscopy reveals ubiquitous presence ofoxidized and reduced quinones in dissolved organic matter. Environmental Scienceand Technology 2005;39:8142–9.

D'Sa EJ, Miller RL. Bio-optical properties in waters influenced by the Mississippi Riverduring low flow conditions. Remote Sensing of Environment 2003;84:538–49.

D'Sa EJ, Miller RL, Del Castillo C. Bio-optical properties and ocean color algorithms forcoastal waters influenced by the Mississippi River during a cold front. AppliedOptics 2006;45(28):7410–28.

Das A, Justic D, Swenson E. Modeling estuarine-shelf exchanges in a deltaic estuary:implications for coastal carbon budgets and hypoxia. Ecological Modelling 2009doi: 10.1016/j.ecolmodel.2009.01.023.

D'Sa EJ. Colored dissolved organic matter in coastal waters influenced by the AtchafalayaRiver, USA: effects of an algal bloom. Journal of Applied Remote Sensing 2008;2:023502.

Feijtel TC, Delaune RD, Patrick Jr WH. Carbon flow in coastal Louisiana. Marine EcologyProgress Series 1985;24:255–60.

Fellman JB, D'Amore DV, Hood E, Boone RD. Fluorescence characteristics and biodegrad-ability of dissolved organic matter in forest and wetland soils from coastal temperatewatersheds in southeast Alaska. Biogeochemistry 2008;88:169–84.

Harvey GR, Boran DA. The geochemistry of humic substances in seawater. In: Aiken GR,McKnight DM,WershawRL,MacCarthy P, editors. Humic substances in soil, sediment,and water. Wiley-Interscience; 1985. p. 233–47.

Holbrook RD, Yen JH, Grizzard TJ. Characterizing natural organic material from theOccoquan Watershed (Northern Virginia, US) using fluorescence spectroscopy andPARAFAC. Science of the Total Environment 2006;361:249–66.

Hua B, Dolan F, McGhee C, Clevenger TE, Deng B. Water-source characterization withfluorescence EEM spectroscopy: PARAFAC analysis. International Journal of Environ-mental Analytical Chemistry 2007;87(2):135–47.

Inoue M, Park D, Justic D, Wiseman Jr WJ. A high-resolution integrated hydrology-hydrodynamic model of the Barataria Basin system. Environmental Modelling andSoftware 2008;23:1122–32.

Jones RF, Baltz DM, Allen RL. Patterns of resource use by fishes and macroinvertebratesin Barataria Bay, Louisiana. Marine Ecology Progress Series 2002;237:271–89.

Kortelainen P, Mattsson T, Laubel A, Evans D, Cauwet G, Räike A. Sources of dissolvedorganic matter from land. In: Sǿndergaard M, Thomas DN, editors. Dissolved OrganicMatter (DOM) in Aquatic Ecosystems: A Study of European Catchments and CoastalWaters. A publication by the EU project DOMAINE (EVK3-CT-2000-00034); 2004.p. 15–22. Chap. 2.

Kowalczuk P, Ston-Egiert J, Cooper WJ, Whitehead RF, Durako MJ. Characterization ofchromophoric dissolved organic matter (CDOM) in the Baltic Sea by excitationemission matrix fluorescence spectroscopy. Marine Chemistry 2005;96:273–92.

Kowalczuk P, Durako MJ, Young H, Kahn AE, Cooper WJ, Gonsior M. Characterization ofdissolved organic matter fluorescence in the South Atlantic Bight with use of PARAFACmodel: Interannual variability. Marine Chemistry 2009 doi: 10.1016/j.marchem.2009.01.015.

Lochmuller CH, Saavedra SS. Conformational changes in a soil fulvic acidmeasured by timedependent fluorescence depolarization. Analytical Chemistry 1986;38:1978–81.

Luciani X, Mounier S, Paraquetti HHM, Redon R, Lucas Y, Bois A, Lacerda LD, RaynaudM,Ripert M. Tracing of dissolved organic matter from the SEPETIBA Bay (Brazil) byPARAFAC analysis of total luminescence matrices. Marine Environmental Research2008;65:148–57.

McKnight DM, Boyer EW, Westerhoff PK, Doran PT, Kulbe T, Andersen DT. Spectrofluo-rometric characterization of dissolved organic matter for indication of precursororganic material and aromaticity. Limnology and Oceanography 2001;46(1):38–48.

Mossa J. Sediment dynamics in the lowermost Mississippi River. Engineering Geology1996;45:457–79.

Nieke B, Reuter R, Heuermann R, Wang H, Babin M, Therriault JC. Light absorption andfluorescence properties of chromophoric dissolved organic matter (CDOM), in theSt. Lawrence estuary (Case 2 waters). Continental Shelf Research 1997;17:235–52.

Ohno T. Fluorescence inner-filtering correction for determining the humification index ofdissolved organic matter. Environmental Science and Technology 2002;36:742–6.

Ohno T, Bro R. Dissolved organic matter characterization using multiway spectraldecomposition of fluorescence landscapes. Soil Science Society of America Journal2006;70:2028–37.

Riu J, Bro R. Jack-knife technique for outlier detection and estimation of standard errorin PARAFAC models. Chemometrics and Intelligent Laboratory Systems 2003;65(1):35–49.

Sampere TP, Bianchi TS, Wakeham SG, Allison MA. Sources of organic matter in surfacesediments of the Louisiana Continental margin: effects of major depositional/transport pathways and Hurricane Ivan. Continental Shelf Research 2008;28:2472–87.

Stedmon CA, Bro R. Characterizing dissolved organic matter fluorescence with parallelfactor analysis: a tutorial. Limnology and Oceanography: Methods 2008;6:572–9.

Stedmon CA, Markager S, Bro R. Tracing dissolved organic matter in aquaticenvironments using a new approach to fluorescence spectroscopy. MarineChemistry 2003;82:239–54.

3221S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222

Author's personal copy

Stedmon CA, Markager S. Resolving the variability in dissolved organic matterfluorescence in a temperate estuary and its catchment using PARAFAC analysis.Limnology and Oceanography 2005a;50(2):686–97.

Stedmon CA, Markager S. Tracing the production and degradation of autochthonousfractions of dissolved organic matter using fluorescence analysis. Limnology andOceanography 2005b;50(5):1415–26.

Stedmon CA, Markager S, Sǿndergaard M, Vang T, Laubel A, Borch NH, Windelin A.Dissolved Organic Matter (DOM) export to a temperate estuary: seasonalvariations and implications of land use. Estuaries and Coasts 2006;29(3):388–400.

Stedmon CA, Thomas DN, Granskog M, Kaartokallio H, Papadimitriou S, Kuosa H.Characteristics of dissolved organic matter in Baltic Coastal sea ice: allochthonousor autochthonous origins? Environmental Science and Technology 2007;41:7273–9.

Swenson EM, Cable JE, Fry B, Justic D, Das A, Snedden G, Swarzenski C. Estuarineflushing times influenced by freshwater diversions. In: Singh VP, Xu YJ, editors.Coastal Hydrology and Process. CO, USA: Water Resource Publications; 2006.p. 403–12.

Tzortziou M, Neale PJ, Osburn CL, Megonigal JP, Maie N, Jaffé R. Tidal marshes as asource of optically and chemically distinctive colored dissolved organic matter inthe Chesapeake Bay. Limnology and Oceanography 2008;53(1):148–59.

Wissel B, Gace A, Fry B. Tracing river influences on phytoplankton dynamics in twoLouisiana estuaries. Ecology 2005;86(10):2751–62.

Yamashita Y, Jaffé R, Maie N, Tanoue E. Assessing the dynamics of dissolved organicmatter (DOM) in coastal environments by excitation emission matrix fluorescenceand parallel factor analysis (EEM-PARAFAC). Limnology and Oceanography2008;53(5):1900–8.

Zepp RG, Sheldon WM, Moran MA. Dissolved organic fluorophores in southeastern UScoastal waters: correction method for eliminating Rayleigh and Raman scatteringpeaks in excitation–emission matrices. Marine Chemistry 2004;89:15–36.

Zhi-gang W, Wen-qing L, Nan-jing Z, Hong-bin L, Yu-jun Z, Wei-cang S, Jian-guo L.Compositional analysis of colored dissolved organic matter in Taihu Lake based onthree dimension excitation-emission fluorescence matrix and PARAFACmodel, andthe potential application in water quality monitoring. Journal of EnvironmentalSciences 2007;19:787–91.

Zsolnay A, Baigar E, Jimenez M, Steinweg B, Saccomandi F. Differentiating withfluorescence spectroscopy the sources of dissolved organic matter in soilssubjected to drying. Chemosphere 1999;38(1):45–50.

3222 S. Singh et al. / Science of the Total Environment 408 (2010) 3211–3222