Organic Geochemistry 58 (2013) 125–136

Contents lists available at SciVerse ScienceDirect

Organic Geochemistry

journal homepage: www.elsevier .com/locate /orggeochem

Diagenetic changes in Lake Superior sediments as seen from FTIR and 2Dcorrelation spectroscopy

Hongyu Li a, Elizabeth C. Minor b,⇑, Prosper K. Zigah a,c

a Large Lakes Observatory and Water Resources Science Program, University of Minnesota, Duluth, MN 55812, USAb Large Lakes Observatory and Department of Chemistry and Biochemistry, University of Minnesota Duluth, Duluth, MN 55812, USAc Department of Surface Waters, Swiss Federal Institute of Aquatic Science and Technology (Eawag), Kastanienbaum, Switzerland

a r t i c l e i n f o

Article history:Received 12 July 2012Received in revised form 23 February 2013Accepted 6 March 2013Available online 15 March 2013

0146-6380/$ - see front matter � 2013 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.orggeochem.2013.03.002

⇑ Corresponding author. Tel.: +1 218 726 7097.E-mail address: eminor@d.umn.edu (E.C. Minor).

a b s t r a c t

Isotopic and elemental analysis, Fourier transform infrared spectroscopy (FTIR), principal componentsanalysis (PCA) and two dimensional (2D) correlation analysis, where core depth was used as perturba-tion, were used to study the diagenesis of organic matter (OM) in Lake Superior sediments. Changes inOM composition were examined at five lake stations over a depth range of 0–10 cm. PCA results showthat depth-related changes among sites are similar, leading to an increased contribution from inorganic(and possibly refractory aromatic organic) components at each site, and a loss of contribution from otherorganic components. Synchronous spectra reveal that aliphatic esters and carbohydrates are degradedsignificantly with increasing depth, leading to an increased contribution from clay/biogenic silica/inactivecarbohydrates. Asynchronous spectra show that, in general, carboxyl groups, including aliphatic ester andamide in protein, are degraded first, followed by a group of carbohydrates and then aromatic compoundsand/or the SiAO framework in clay and biogenic silica. Site dependent compositional variation occurs andappears to be influenced by topography and geology, e.g. the delivery of a larger load of terrestrial inor-ganic silicate minerals to certain sites and re-suspension/re-deposition, leading to less intensive downcore variation at mid-lake central and eastern basin sites. The study demonstrates the usefulness of FTIRcoupled with PCA and 2D correlation approaches for exploring structural changes in sedimentary mate-rial during diagenesis.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Early diagenesis of primary organic macromolecules in theoceans is often conceptualized as rapid microbial and chemicaldestruction of the biomacromolecules (mainly in the water col-umn) followed by slower condensation of the metabolic andhydrolysis products to form refractory humic material, which takesplace mainly in ocean sediments (Degens, 1967; Kemp andJohnston, 1979). In the Great Lakes, which are much shallower,the role of the sediment–water interface as an area of organic com-ponent destruction is likely to be enhanced. In Lakes Ontario andErie, the most intense microbial activity and the greatest bacterialconcentration have been found at the sediment–water interface(Vanderpost, 1972; Dutka et al., 1974). Downcore diagenetic alter-ation of some important paleo-proxies including bulk stablecarbon isotopic, stable nitrogen isotopic and C/N ratio, as well asthe concentration of humic substances and some important lipidbiomarkers in Great Lakes sediments have been studied (reviewedby Meyers and Ishiwatari (1993)). Also, Kemp and Johnston (1979)used wet chemistry analysis to investigate the diagenesis of

ll rights reserved.

extractable OM in sediments from Lakes Ontario, Erie and Huron.They identified trophic state and water depth as the two controlson the degree of diagenesis of modern sedimentary OM. They alsofound that the decomposition rate was in the following order: aminoacids� amino sugars > carbohydrates > humic compounds > lipids.Although there have been studies of individual paleo-proxies ofselected extractable fractions like hydrolysable amino acids, aminosugars and carbohydrates, little is known about the composition andtransformation of the high molecular weight (MW) entities (bothprecursors of the extractable material and the portions not amenableto extraction).

With the application of 2D correlation approaches to FTIR data(Abdulla et al., 2010b), more structural information can be ex-tracted from the spectra of complex mixtures such as those in sed-iments. Noda (1993) developed the concept of generalizedperturbation based 2D correlation spectroscopy, which works byspreading the spectral intensity changes within a data set collectedacross a perturbation (time, temperature, depth) across a seconddimension as a function of the perturbation. As pointed out byNoda and Ozaki (2004), by applying 2D correlation spectroscopy,spectral resolution is enhanced by spreading peaks over a seconddimension; the correlations among IR bands facilitate functionalgroup assignments, and the specific sequential order of spectral

126 H. Li et al. / Organic Geochemistry 58 (2013) 125–136

intensity changes through a data set can be identified throughasynchronous analysis. The use of 2D spectroscopy is resolvingmany ambiguities in natural complex mixtures of compounds withmany overlapping peaks (Noda and Ozaki, 2004; Mecozzi et al.2009a,b,c; Abdulla et al., 2010a,b; Popescu et al., 2010). Using 2DFTIR in conjunction with multivariate statistical methods (suchas discriminant analysis, PCA and cluster analysis) can help con-firm trends within an FTIR data set and provide additional comple-mentary insight into the distribution of sample compositions.

In this study, isotopic and elemental analysis, FTIR, PCA and 2Dcorrelation analysis, with core depth as the perturbation, are usedto study diagenesis in sediments from Lake Superior. SedimentaryOM is characterized in terms of the functional groups of the prin-cipal building blocks of living aquatic organisms (e.g. proteins, car-bohydrates and lipids).

2. Material and methods

2.1. Study site and sampling

Lake Superior is the Earth’s largest lake by surface area(82,100 km2), with a maximum depth > 400 m, mean depth150 m and water residence time 180 yr; it is dimictic, mixing twiceannually (Assel, 1986; Herdendorf, 1990; Urban et al., 2005). Or-ganic carbon (OC) deposition in the sediments is ca. 0.48 � 1012 -g C yr�1 (McManus et al., 2003) to 1.5 � 1012 g C yr�1 (Johnsonet al., 1982; Klump et al., 1989). Baker et al. (1991) found that < 5%of the OC in surface particulate OC (POC) is transported to the bot-tom to accumulate in the sediment. Offshore areas have a sedimen-tation rate (0.01–0.05 cm yr�1) close to that of ocean environmentsand there is minimum bioturbation in the sediment (Johnson andEisenreich, 1979; Evans et al., 1981; Krezoski, 1989; Li et al.,2012). Surface sediments are oxic, with an average O2 penetration

Fig. 1. Map of Lake Superior. Sampling sites (CM, EM, NM, SM and WM) are shown witriangles were obtained with 210Pb method and are from Evans (1981); up triangles areet al. (1975) and were obtained from 210Pb and pollen methods.

depth of ca. 7 cm at the mid-lake sites (Li et al., 2012). A lack oflimestone in the drainage basin results in remarkably low CaCO3

in the sediment (Matheson and Munawar, 1978).Multi-cores were collected in May 2010 from five stations

(Fig. 1) using an Ocean Instrument multi-corer. The water depthat stations CM, EM, NM, SM and WM was 258 m, 243 m, 218 m,388 m and 171 m, respectively. From each sampling site one ormore cores were extruded and sliced into 0–2 cm, 2–4 cm, 4–6 cm, 6–8 cm and 8–10 cm layers from the sediment–water inter-face. The slices were put into acid cleaned and combusted(450 �C, P 4 h) glass jars and stored at �20 �C. Samples werefreeze-dried, homogenized and stored at room temperature untilfurther analysis.

2.2. Elemental and isotopic analysis

Acid fumigation (Harris et al., 2001) was used to remove car-bonate carbon before elemental analysis (EA) and isotope-ratiomass spectrometry (IR-MS). Weighed samples were placed in cleanAg capsules and 50 ul MilliQ water was added to each sample.Samples were then fumed over 12 M HCl (6–8 h), oven-dried(60 �C, 4 h) and cooled. For more complete combustion, each sam-ple (in the Ag capsule) was enclosed within a Sn capsule for %C, %N,d13C and d15N determination with a Costech CNS analyzer inter-faced with a ThermoFinnigan Delta-PlusXP stable isotope ratiomass spectrometer.

Radiocarbon measurements were performed at the NationalOcean Sciences Accelerator Mass Spectrometry Facility (NOSAMS)at the Woods Hole Oceanographic Institution. Homogenized, acidtreated and dried samples were combusted to CO2, reduced tographite and analyzed using accelerator mass spectrometry todetermine radiocarbon content. Values are reported as D14Caccording to the convention of Stuiver and Polach (1977).

th stars. Sedimentation rate from previous studies is included for reference: Downfrom Kemp and Johnston, 1979.) using the pollen method; circles are from Bruland

H. Li et al. / Organic Geochemistry 58 (2013) 125–136 127

2.3. FTIR spectral measurement and pre-processing

Spectra were obtained with a Nicolet Magna IR 560 FTIR ESPspectrometer equipped with purge gas generator unit by collecting200 scans from 4000 to 400 cm�1 with resolution 2 cm�1 and usingHapp-Genzel apodization. For expelling CO2 contamination fromair, 1 lag time (5 min) between closing the analytical chamberand starting analysis was maintained. Freeze-dried, homogenized(but not acid-treated) samples were run as KBr pellets (1 mg sam-ple, 130 mg KBr, homogenized in a mortar and pestle, pressed at1500 psi). The pellet was dried in a desiccator ( > 6 h) before anal-ysis. A pure KBr pellet was analyzed before each sample analysis asa background blank. Spectra were converted to absorbance unitsand baseline corrected. The second derivatives of the spectra weregenerated using Origin 7.0 and the 2nd order Savitzky–Golaymethod with 11 convolution points.

For comparison with bulk freeze-dried sediment, an additionalaliquot of the same sample was combusted at 400 �C for 8 h (assuggested by Ball (1964) and Ben-Dor and Banin (1989)) to removeOM while preserving inorganic components as far as possible.Comparing FTIR spectra from combusted and uncombusted ali-quots assisted in the identification of key inorganic vs. organic con-stituents in the samples.

2.4. 2D correlation and PCA

Before 2D correlation analysis, each spectrum was loaded intoMATLAB 7.0 and area-normalized (Abdulla et al., 2010a) to mini-mize the influence of variable sample loading on the correlationanalysis. The PCA noise reduction method proposed by Jung(2004) and applied by Abdulla et al. (2010b) was used to reducenoise prior to 2D correlation analyses. 2D FTIR correlation spectrawere constructed for each of the five sites with core depth (layer)as the perturbation, sorting from the surface sediment layer todepth, in order to study the change in OM composition over thisrange. The area-normalized, noise-free data sets were importedinto a modified version of 2Dshige software (Kwansei-Gakuin

0 2 4 60

5

10(A) C concentration (%)

Dep

th (m

)

-32 -30 -28 -26 -240

5

10

(C) δ13 C ( ‰ )

Dep

th (m

)

EMCMNMWMSM

Fig. 2. Elemental and isotopic analysis results from Lake Superior sediments, showing do

University, Japan, available at http://sci-tech.ksc.kwansei.ac.jp/~ozaki/) to generate 2D-correlation spectra.

Besides its use as a noise reduction method, PCA was also usedin its more traditional sense for classifying samples, in order to ex-plore the compositional differences and similarities among sam-ples. For classification, the data matrix was z-scored and variancescaled before the PCA analysis was performed. This minimizesthe influence of varying response factors (band intensity differ-ences due to the physical variation in different functional groups)on the PCA groupings. Matlab 7.0 and in-house code were usedfor processing (Stephens and Minor, 2010).

3. Results and discussion

3.1. Elemental and isotopic analysis

The origin, composition and diagenesis of the OM were initiallyassessed using C and N concentrations and d13C values (Fig. 2).Consistent downcore d13C values (�25.6‰ to �27.9‰) were seenfor all stations except SM (d13C value �31.5‰ at 2–4 cm). C/Nvalues ranged between 9.7 and 11.6 (except SM 2–4 cm, C/N 8.2).Both d13C and C/N values are similar to reported values (�26.9‰

to �23.9‰ and 9.1–13.4, respectively) for sediments from thewestern basin and off the Keweenaw Peninsula (Ostrom et al.,1998; Urban et al., 2004) and may indicate that the OM in thesediment is derived from autochthonous productivity (Meyersand Ishiwatari, 1993; Ostrom et al., 1998). However, the C/Nvalues are somewhat higher than reported for Lake Superiorphytoplankton, and both the C/N and d13C values are also withinthe range reported for soils (Hoffman et al., 2010). At all sites exceptSM, which appears unique in its elemental and isotopic composition,C/N values show an initial increase with depth followed by a decreasein the deeper sections. This pattern may represent a slight diageneticpreference for the loss of N vs. C at the very beginning, followed by anincrease in the proportion of remaining OM from microbial biomass(more enriched in N-containing compounds, Lehmann et al., 2002)in the deeper part of the core.

0 0.1 0.2 0.3 0.40

5

10(B) N concentration (%)

Dep

th (m

)

8 9 10 11 120

5

10

(D) C/N

Dep

th (m

)

wn core trends in (a) C% (wt./wt.), (b) N% (wt./wt.), (c) d13C and (d) atomic C/N ratio.

128 H. Li et al. / Organic Geochemistry 58 (2013) 125–136

In terms of %C, %N and d13C, downcore trends for EM, CM andNM are similar, with deep site SM and western-lake site WMexhibiting more extreme changes with depth. The trends indicatesite-dependent variation in diagenesis, perhaps related to watercolumn depth and depositional environment (further discussedin Section 3.4.1). While we only have downcore radiocarbon datafor two sites (WM and EM), the downcore differences in OM ageare more extreme for site WM (Table 1), which exhibited the stron-gest downcore decreases in %C and %N (Fig. 2a and b). Radiocarbondata and %C in surface sediments do not, however, appear corre-lated on the basis of Pearson’s linear correlation analysis (N = 5,R2 0.0233, P 0.8064) and may instead reflect varying input fromsources with different pre-aged OC content.

3.2. Insights into sediment OM composition based on 1D FTIRspectroscopy and 2nd derivative spectra

The 1D FTIR spectra from all depths and sites (e.g. Fig. 3) aresimilar. As reported for dissolved OM samples from an ElizabethRiver/Chesapeake Bay salinity transect (Abdulla et al., 2010a), theheterogeneity of the samples actually acts to simplify their FTIRspectra, which exhibit broad bands from the overlap of multiplefunctional groups. Such broadening is also seen in clay analysisusing FTIR (e.g. Madejová and Komadel, 2001). However, even inFig. 3, with samples containing a mixture of inorganic and organic

Table 1D14C and corresponding age (in radiocarbon yr) of sedimentary particulate matterwith porewater removed by centrifuging before freeze-drying.a

Station Sedimentdepth (cm)

Water columndepth (m)

D14C (‰) Age(radiocarbon yr BP)

CM 0–2 258 �19.6 100 ± 20NM 0–2 218 �22.6 125 ± 20SM 0–2 388 �36.6 240 ± 15EM 0–2 243 �35.7 235 ± 15EM 4–6 �126.2 1030 ± 20EM 8–10 �179.5 1530 ± 15WM 0–2 171 �116.8 940 ± 15WM 4–6 �595.8 7220 ± 35WM 8–10 �620.3 7720 ± 35

a Precision is based on error of standards or multiple analyses.

Fig. 3. 1D FTIR spectra of sediments at different depths at (a) CM and (b) WM stations. Eareduce variation from sample loading/processing.

compounds, some peak shift and peak intensity differences can beseen as a function of increasing sediment depth. We therefore ap-plied the 2nd derivative of the FTIR spectra (e.g. Fig. 4) to clarifychanges in the slope in the original spectra, and thus to identifythe number of overlapping bands and the exact frequency of peakresponse (Smith, 1996; Griffiths and De Haseth, 2007).

To assist in identifying inorganic vs. organic components, 1DFTIR and 2nd derivative spectra were compared on aliquots ofthe same sediment sample pre and post combustion.

The comparison suggests that clay/biogenic silica/carbohy-drates are the dominant chemical constituents of Lake Superiorsurface sediment (Fig. 4 and Table 2). Although optimized combus-tion conditions (400 �C, 8 h as suggested by Ball, 1964; Ben-Dorand Banin, 1989) were applied to minimize the loss of structuralwater and ensure complete removal of OC, we are aware thatany degree of dehydration of clay from combustion might lead tochanges in the FTIR spectra, especially the bands related to crystal-lized/absorbed water. However, comparison of the bulk sedimentand combusted sediment spectra (e.g. Fig. 4a and b) still showsthe strong signal from clay minerals (mainly bands from 1000 to1200 cm�1) and recognizable signals from organic compounds(mainly bands from 1200 to 1800 cm�1).

Examination of the 1D and 2nd derivative spectra from all fivestations and all sediment depths (Table 2) shows that all sites ex-hibit strong signals from clay minerals (and possibly some biogenicsilica) and recognizable signals from the functional groups withincarbohydrates, carboxylic acids, aliphatic/acetyl esters, amides/proteins and phenols/lignin. The relative intensity of the 1D peaksis a function of both concentration and response factor. For exam-ple, bands indicating lignin are fairly weak even for measurementsof lignin isolates, thus making it difficult to apply them in a quan-titative fashion to environmental samples.

3.3. PCA of the FTIR data

PCA, a commonly used feature-based classification method(Kvalheim et al., 1985; Sanni et al., 2002), was performed on thedata set comprising normalized FTIR spectra from all stations andcore depths to investigate the variation among samples at differentdepth from different sites. The data matrix was z-scored and vari-ance scaled before the PCA analysis was performed. PC1 and PC2

ch spectrum was normalized via total area as described in text before comparison to

Fig. 4. 1D FTIR spectrum and 2nd derivative for the top 2 cm of the (a) bulk sediment and (b) combusted sediment at the CM station. The combustion conditions (400 �C, 8 h),as suggested by Ball, 1964; Ben-Dor and Banin, 1989 were applied to minimize loss of structural water and ensure complete removal of OC.

H. Li et al. / Organic Geochemistry 58 (2013) 125–136 129

account for 79% of the variability in the data set (Fig. 5a). For all thestations, the top two to three layers tend to be clustered in nega-tive PC1 space. With increasing depth the samples shift towardmore positive PC1 values and the deepest layers are all in positivePC1 space. Positive correlation can be found between PC1 scoreand radiocarbon age for samples from the same station (e.g.N = 3, R2 0.9789, P 0.0929 for WM), with older layers having morepositive PC1 values (data not shown). However, no correlation isfound when all the radiocarbon data from Table 1 are comparedwith FTIR-based PC scores. This is likely due to the lack ofdowncore radiocarbon values and the differences in pre-aged OMsources in the surface sediment layers at different stations.

In PCA, squaring a loading gives an estimate of the leverage eachvariable has on the resulting PC. Variables with large leverage canbe identified from loading plots (e.g. Fig. 5b, a reconstructed load-

ing plot showing loading scaled by instrument response at thatwavelength). Plotting the PC1 loadings (Fig. 5b) illustrates thatsamples located on the positive side (Fig. 5a) have more refractorycharacter (peaks from 3000 to 3500 cm�1 and peaks centered at600 cm�1, consistent with an enhancement in the proportion ofhydrated clay, and perhaps also aromatic organic compounds, inthe sample). In contrast, loadings to the negative side of PC1 in-clude the region from 1200 to 1800 cm�1, consistent with carbox-ylates, aliphatic esters and amide/protein and the region from 1000to 1200 cm�1, which we attribute to carbohydrates as the clay andsilicate peaks in that region should remain relatively stable andthus be subtracted out in the z-scoring step prior to PCA (the rela-tionships in Fig. 5a and b also show that, with labile compoundsbeing degraded via diagenesis, the remaining FTIR-amenable mate-rial (possibly clay, resistant forms of the original organic mixture

Table 2Bands for compound identification in Lake Superior sediment (surface layer) with FTIR 2nd derivative spectra.

Compound class Modek Representative band position (cm�1)

CM EM NM SM WM

Carboxylic acida,b,c,d C@O stretching in COOH (1710 ± 15), + 1712 1716 1710 1716 1717Asym. stretching of ACOOA (1600 ± 40), + 1612 1577 1633 1619 1617

1630Sym. stretching of COOA (1405 ± 45), + 1414 1406 1413 1400 1407OAH in-plane bending (1418 ± 22),l + 1426 1418 1432 1434 1417

Amidea,d,e,f Amide I C@O stretching (1650 ± 20), ++ 1649 1646 1633 1652 16531658 1655 1643 1662

Amide II C@O stretching (1550 ± 20), + 1554 1540 1551 1542 1559

Estera,d Aliphatic, C@O stretching (1740 ± 10), + 1739 1734 1730 1748 1734Aliphatic, CACAO stretching (1185 ± 25), +++ 1176 1175 1177 1176 1174Acetyl, CACAO stretching (1240 ± 5), ++ 1238 1237 1238 1237 1240

Ligninh,i,j Vibrations of aromatic rings (1515 ± 5), + 1515 1515 1512 1520 1519

Carbohydrated and clay/biogenic silicag OAH stretching �3400,l ++ 3399 3405 3400 3388 3392SiAOH stretching, ++ 3697 3700 3697 3699 3700

3621 3620 3621 3620 3621CAO asym. stretching (1100 ± 100),l +++ 913 913 913 913 913

1007 1008 1007 1007 10041038 1038 1039 1037 10371095 1095 1095 1079 1096

or SiAO asym. stretching (1000 ± 100), +++ 1165 1164 1165 1164 1165

g Pandurangi et al. (1990)k Relative response in 1D spectrum is indicated with +++ for intense, ++ for moderate, and + for weak; note that the relative response was similar for all samples.a Smith (1999)b MacCarthy and Rice (1985)c Celi et al. (1997)d Silverstein and Webster (1997).e Harisa and Severcanb (1999)f Widjanarko et al. (2011)

h Mascarenhas and Arbuckle (2000)i Raiskila et al. (2007)j Derkacheva and Sukhov (2008)l Signal at 1643 cm�1 could be from HAOAH bending in water adsorbed to clay (e.g., Madejová, 2003); 1165 cm�1 is from quartz (Ramachandran and Beaudoin, 2008); very

broad band around 3400 cm�1 can be attributed to overlap of OH groups bonded to clay, OAH stretching of phenol, carbohydrate and carboxylic acid compounds and NAHband in amide; CAO asym. stretching from carbohydrates and SiAO asym. stretching from clay/biogenic silica overlaps at 1000–1100 cm�1 range.

Fig. 5. PCA projection (a) and corresponding PC1 loading plot (b) for normalized FTIR data. For Fig. 5a, samples are named as the first station letter (see Table 1) plus a numberindicating depth in sediment core: 1, 0–2 cm; 2, 2–4 cm; 3, 4–6 cm; 4, 6–8 cm; 5, 8–10 cm. Deepest layers are shown with bold italic font and underlining.

130 H. Li et al. / Organic Geochemistry 58 (2013) 125–136

and diagenetic products) becomes more similar across the lakesites.

There were interesting site-specific variations. SM showed thestrongest variation in composition as a function of depth. WMsamples from all core depths plotted closer to the ‘deep layergroup’, while NM samples appeared less diagenetically altered, asindicated by the top four layers being outside the ‘deep layergroup’. The FTIR PCA-based diagenetic state of the WM sedimentsagrees with both our 2D correlation FTIR results (see below) andour radiocarbon ages (Table 1).

3.4. Diagenetic change with depth as revealed from 2D correlationFTIR

3.4.1. 2D FTIR synchronous map2D correlation spectroscopy was performed according to Noda

and Ozaki (2004) as described by Abdulla et al., 2010b to extractcompositional changes in sediment with depth. Due to its mathe-matical nature, perturbation based 2D correlation analysis is espe-cially sensitive to noise. The noise can be the high frequency noiseassociated with the instrument’s detector and electronic circuits; it

H. Li et al. / Organic Geochemistry 58 (2013) 125–136 131

can also be low frequency noise and localized noise, such as lowfrequency noise caused by instrument drift during scanning. Noisein the FTIR data could cause artifacts in the 2D treatment, especiallyfor asynchronous maps (Noda and Ozaki, 2004), so we applied thePCA noise reduction method proposed by Jung (2004) and used byAbdulla et al. (2010b) as the noise filter. The first two PCs contained> 92% of the variation in each data set; reconstructed data sets usingthese two PCs were the starting point for 2D correlation studies. Theremoved spectra appear to consist mainly of noise, exhibiting norecognizable features (Supplementary Fig. S-1). To further checkthat the approach was appropriate, we compared synchronousand asynchronous contour maps with and without noise, as recom-mended by Abdulla et al. (2010b). As in that study, our synchronousmap shows no noticeable differences in the presence or absence ofnoise; the asynchronous contour maps show some artifacts whenusing the original noise-containing spectra (data not shown). All2D correlation results were generated from the noise free spectra.

Spectral data were constructed for each site with core depth asthe perturbation, sorting from the surface sediment layer to depth.Since 210Pb geochronology studies revealed that the offshore areaof Lake Superior has a low sedimentation rate (0.01–0.05 cm yr�1

minimum influence from wind-generated re-suspension and min-imum bioturbation of only 1–2 cm in the sediment (Johnson andEisenreich, 1979; Evans et al., 1981; Krezoski, 1989; Li et al.,2012), diagenesis is believed to be the main reason for OM compo-sition changes vs. depth.

The synchronous map shows a symmetric spectrum with re-spect to the diagonal line. Major auto-peaks spread along the diag-onal line represent correlation between a band and itself along acertain perturbation. For example, here the auto-peaks representvariation in FTIR response (‘‘concentration’’) as a function ofincreasing depth; the intensity of an auto-peak represents theintensity of the variation. Thus, 2D correlation analysis has theability to extract only the FTIR responses that change through per-turbation out of a complex background, such as a strong stableclay/refractory OM background in the sediments. Note that all autopeaks are positive by definition, indicating a correlation; thesecorrelations could represent either an increase or a decrease inconcentration with depth; the direction of change has to beconfirmed by comparison of the original spectra.

Fig. 6. Synchronous 2D correlation spectrum down core for CM station. The spectral datawas normalized and PCA reconstructed before 2D correlation analysis (see text for detailsthe right of each 2D correlation spectrum. White indicates positive correlation, black ne

Taking CM (Fig. 6) as an example, 13 distinct auto peaks are ob-served in the 2D synchronous correlation spectrum. The greatestchange in intensity lies in the multiple bands around 1000–1100 cm�1, consistent with SiAO stretching in clay minerals andbiogenic silica or CAO stretching in carbohydrates. Consideringthe general stability of the SiAO bond in clay minerals/biogenic sil-ica and as evidenced by the disappearance of the main SiAO claypeaks (1065 cm�1, 1095 cm�1, 1037 cm�1) in 2D correlation spec-tra, any losses indicated by these auto peaks should result mainlyfrom the diagenesis of labile carbohydrates with a possible minorcontribution from biogenic silica dissolution. The band showingthe second-greatest change in intensity (ca. 1193 cm�1) is attrib-uted to aliphatic ester or acetyl ester. The band at 1659 cm�1

(amide C@O asymmetric stretching) is small, but implies that pro-tein/peptide content changes with diagenesis. The broad auto peakaround 3344 cm�1 (–OH stretch region) indicates changing of phe-nol, carbohydrate, long chain alcohol and/or carboxylic acid groupsover depth. Absorbance and desorption of water to clay minerals,and thus a change in clay mineralogy with depth, could also con-tribute to this band. The peak around 3642 cm�1 could be a silanolSiAOH stretch and may indicate the early stages of biogenic silicadiagenesis. Thus, the 2D correlation map (Fig. 6) implies that in thetop 10 cm of sediment, carbohydrates and aliphatic esters changeto the greatest extent, while changes in amide/protein, some com-bination of phenol/carbohydrate/carboxylic acid/clay minerals(water absorbance and desorption), as well as biogenic silica, alsooccur. The changes in carbohydrate (1092 cm�1), aliphatic ester,amide/protein, OAH groups and silanol SiAOH in biogenic silicawere confirmed as a decrease by finding depth-related decreasesin the peak intensities of the corresponding wavenumbers in thearea-normalized 1D FTIR (data not shown). The clay/biogenic sil-ica/carbohydrate band at 1003 cm�1 is confirmed as an increase,which is likely due to the significant loss of other compounds dur-ing digenesis, leaving some clay and possibly some non-active car-bohydrate/biogenic silica as a greater proportion of total sediment.

In order to identify how these major bands are correlated witheach other, the off-diagonal peaks (cross-peaks) in the synchro-nous map were investigated. Bands represented by X and Y coordi-nates of positive cross-peaks share the same directional variationwith depth and the relative intensity changes correspond to the

set was constructed sorting from the shallowest depth to the deepest. The data set). For ease of identification of peaks the slice through auto peak diagonal is plotted togative correlation.

132 H. Li et al. / Organic Geochemistry 58 (2013) 125–136

relative strength of the change in FTIR response of the functionalgroups. For CM (Fig. 6), two sets of functional groups are identifiedthat respond differently to depth. The first set is identified by thepositive cross-peaks among SiAOH (3642 cm�1), amide/protein(1659 cm�1 and 1530 cm�1), aliphatic ester (1193 cm�1) and car-bohydrate/silicate (1092 cm�1); comparison of original spectrashows that this shift is a decrease in proportional response withdepth in the sediment. The second set is identified by positivecross-peaks among the clay/biogenic silica/carbohydrate band(1003 cm�1), unsaturated CAH (3133 cm�1), methyl CAH (2884and 2836 cm�1) and the bonds represented by 862 cm�1,593 cm�1 and 536 cm�1; these FTIR bands increase with depth,as indicated by their negative cross-peak relationship with the firstset of functional groups (Fig. 6). Based on the above cross-peakanalysis, peaks at 1092 cm�1 and 1003 cm�1 represent differentcarbohydrate and or silicate moieties varying significantly in dia-genetic response. The negative correlation of band 1003 cm�1 vs.other functional groups (and its increasing proportion with depth)here is most likely due to the inactive clay minerals and biogenicsilica species dominating the signal at this range, and thus is re-ferred to here as clay/biogenic silica/carbohydrate band.

Although it is hard to assign specific vibration modes to wave-numbers below 1000 cm�1, peaks centered around 800 and600 cm�1 can be attributed to SiO vibrations (quartz, Herbert et al.,1992; Vogel et al., 2008) and to signals from Mg3OH in chrysotile(Madejová, 2003) and from montmorillonite (620 cm�1, Madejováand Komadel, 2001); these peaks also coincide with those attributedto aromatic vibrations. The peaks show a positive correlation withthe clay/biogenic silica/carbohydrate band (1003 cm�1), unsatu-rated CAH (3133 cm�1) and methyl CAH (2884 and 2836 cm�1),and a negative correlation with other functional groups (Fig. 6). Dia-genesis over depth in sediment leads, therefore, to an increase in therelative proportion of clays/biogenic silica and methylated, unsatu-rated organic and possibly aromatic compounds relative to otherfunctional groups in the sedimentary OM.

The absence of cross-peaks and auto peaks in bands such as1700 and 1640 cm�1 in 2D FTIR spectra indicates an absence/scar-

Table 3Compound class correlation relationship (varying with depth) as indicated from 2D synch

Compound classa (CM) A B C D E F GA (1193) + + � � + �B (1659, 1530) + + � � + �C (1092) + + � � + �D (1003) � � � + � +E (3344) � � � + � +F (3642) + + + � � �G (3133, 539) � � � + + �

Compound classa (NM) A B C D E F GA (1250, 1730) N � � + N +B (none) N N N N N NC (1087) � N + � N �D (1031) � N + � N �E (3402) + N � � N +F (none) N N N N N NG (622, 567) + N � � + N

Compound classa (WM) A B C D E F GA (1210) + N � N N �B (1653, 1559) + N � N N �C (none) N N N N N ND (1012) � � N N N +E (none) N N N N N NF (none) N N N N N NG (531) � � N + N N

a A, aliphatic ester; B, amide/protein; C, carbohydrate; D, clay/biogenic silica/carbocompound.

b N indicates no identifiable cross-peak.

city of carboxylic acid diagenesis. Since we do find carboxylic acidsignals in 1D FTIR, it appears that carboxylic acids are relativelyconservative in the sediments. This behavior may be because themore labile portions (such as unsaturated fatty acids) have alreadybeen reworked in the water column or because the alteration offatty acid structures and distributions is too subtle to be seen byFTIR.

As seen in Table 3 (and Supplementary, Fig. S-2) and in theasynchronous study below in Section 3.4.2, the activity of OAH/NAH bands is site dependent. A combination of polymeric alcohol,carboxylic acid, amide and even water absorbed to clay minerals/silica as discussed above, could contribute to these bands, andthe varying contributions from these compound classes at eachstation most likely leads to the OAH/NAH activity difference.OAH/NAH bands were found to be positively correlated with inac-tive clay/carbohydrate bands around 1000–1100 cm�1 and nega-tively correlated with other functional groups at SM and CM, thetwo deepest water sites (388 and 258 m, respectively), while theycorrelated positively with carbohydrate bands, aliphatic ester andamide/protein at other stations (whose water depth ranged from171 to 243 m).

NM, SM and WM sediments (Table 3 and Supplementary, Fig. S-2b–d) appear to have a stronger and somewhat different diageneticresponse vs. CM and EM. For all three of these stations, the corre-lation coefficient for the auto peaks is considerably higher thanseen for CM and EM. First of all, the clay/biogenic silica/carbohy-drate bands show negative correlation with other functionalgroups in the 1100–2000 cm�1 range, including aliphatic esterbands at all the stations and the amide/protein bands at WM andSM. At NM and WM, the intensity of SiAOH auto peaks andcross-peaks (around 3629 cm�1) is very low vs. other stations.The diagenetic differences at NM, SM and WM vs. CM and EM couldbe caused by variation in the sediment source and age, sedimenta-tion rate (Fig. 1), water column depth and depositional environ-ment at the sites.

NM differs from all the other sites in its lack of a positive rela-tionship between the aliphatic ester bands and protein, and its po-

ronous FTIR cross-peaks.

Compound classa (EM) A B C D E F GA (1217, 1725) + + � + � �B (1658, 1537) + + � + Nb �C (1087, 1040) + + � + � �D (988) � � � � + +E (3404) + + + � � �F (3629) N N � + � +G (586) � � � + � +

Compound classa (SM) A B C D E F GA (1224) + N � � � �B (1544) + N � � � �C (none) N N N N N ND (1013) � � N + + +E (3427) � � N + + +F (3621) � � N + + +G (590) � � N + + +

hydrate; E, OH/NH groups; F, biogenic silica/silicate; G, clay/inorganic/aromatic

H. Li et al. / Organic Geochemistry 58 (2013) 125–136 133

sitive correlation of the aromatic band and aliphatic esters(Table 3). It also differs from the other sites in lacking a labile pro-tein signature (there is no auto peak for protein bands at this site;Table 3 and Supplementary, S-2b). These trends indicate increasedterrestrial, rather than autochthonous, input at this site, consistentwith circulation patterns in the lake (Bennington et al., 2010). Withdiagenesis, the proportion of relatively inactive compounds wouldincrease rapidly as the labile compounds were depleted, resultingin the strong carbohydrate/silicate signal being negatively corre-lated with other bands (Table 3 and Supplementary Fig. S-2).

SM has the most distinct 2D synchronous spectrum, with a ca.10� greater correlation coefficient than all other stations (asshown by y-axis in the auto spectra in Fig. 6 and SupplementaryFig. S-2), indicating much more intensive variation in compoundcomposition at this site. Core SM was recovered from the deepestlocation (Fig. 1), where bottom currents are accelerated and ero-sional features are common (Johnson, 1980). In addition to its dif-ferences in FTIR parameters, it also exhibits significantly lower %Cand %N relative to the other sites. These differences are likely dueto the significantly stronger bottom current scouring away lighterOM-rich sediment. Thus there is most likely a fine layer of very re-cent organic material that was just deposited (and not yet scouredaway) underlain by relatively coarse grained material with highlyreworked, refractory OM. Evidence for this is found in the physicalproperties of this core relative to others: SM has coarser sedimentrelative to the other cores and SM sediments are yellow (vs. olivegray at other stations), indicating less OM-rich material.

The diagenetic differences at WM are more difficult to explain.WM exhibits stronger changes in %C and %N with depth than NM,CM and EM (Fig. 2) and may be more impacted by the generallyhigher primary productivity in the western arm of the lake(McManus et al. 2003). However, its surface sediment (0–2 cm) isthe oldest of the sample sites (Table 1) and may indicate additionalsources of pre-aged material at this location.

Another possible scenario to explain the differences in WM, NMand SM vs. EM and CM is the intensive re-suspension believed tooccur near EM and CM, as reported by Zigah et al. (2012) on the ba-sis of radiocarbon distributions in lake particulate OM at thesesites. Such re-suspension events could shift much of the OM dia-genesis into the water column rather than surface sediment,although if this were the case, we would expect to see lower %Cin surface sediment at these sites, rather than the somewhat higherproportion actually measured (Fig. 2). A second possibility is thatsites EM and CM are areas of deposition for resuspended surfacesediment transported from elsewhere in the lake, and thus exhibitless variation down core than sites more impacted by surfacewater primary production. Without the intense decomposing of athin layer of fresh OM at the sediment interface, the clay/biogenicsilica/carbohydrate region will not show a negative correlationwith other functional groups as seen at WM, NM and SM. This sce-nario is consistent with the correlation coefficients of the autospectra (Fig. 6 and Supplementary Fig. S-2); EM and CM have a2D correlation coefficient roughly 7–30� smaller than the otherstations, indicating much less intense compositional variation inthe functional groups, especially the clay/biogenic silica/carbohy-drate region, as a function of depth. It is also likely that the OMcomposition of each site is influenced by a varying combinationof terrestrial OM/inorganic matter delivery vs. delivery of primaryproduction, as well as differences in transport and depositionalprocesses. The interaction of these processes would be a usefularea for further research.

3.4.2. 2D FTIR asynchronous mapThe asynchronous map supports and extends observations from

the synchronous map and PCA analysis. An asynchronous cross-peak between a pair of FTIR bands develops only if the intensities

of these two spectral features change out of phase with each other,and thus in our application, show a different depth scale of diage-netic response. Noda’s rule (Noda and Ozaki, 2004) can be appliedto the asynchronous data in order to determine the sequential or-der of the changes between pairs of FTIR bands. The rule basicallystates that, assuming that features x and y are positively correlatedin the synchronous spectrum (see Table 3), in an asynchronousspectrum, a positive cross-peak at x and y indicates a change in fea-ture x before the change in feature y. For the same circumstances, anegative cross-peak at x, y indicates that feature x’s change lags be-hind that of feature y. This sign rule is reversed when the synchro-nous correlation intensity at the same coordinates is negative(again see Table 3). Because through diagenesis the variation pat-terns of different functional groups in Lake Superior sediment arepresumably monotonic, Noda’s rule is reliable and is used here toidentify the relative order of response of different functionalgroups to diagenesis.

For example, CM (Fig. 7a), exhibits positive asynchronous cross-peaks for the following species on the x-axis relative to a y of1003 cm�1: 3633 cm�1, 1658 cm�1, 1537 cm�1, 1193 cm�1,1092 cm�1. These, coupled with the presence of negative synchro-nous peaks at the same coordinates (Table 3 and SupplementaryFig. S-3), indicate that the intensity of the clay/biogenic silica/car-bohydrate band at 1003 cm�1 changes predominantly before thechange in amide/protein (1658 cm�1, 1537 cm�1), aliphatic ester(1193 cm�1), another group of carbohydrates (1092 cm�1) andSiAOAH bands in biogenic silica (3633 cm�1). It was shown abovethat this clay/biogenic silica/carbohydrate band increases in rela-tive intensity with increasing depth, most likely indicating its per-sistence as other species are removed via diagenesis (although itsown generation within the sediments cannot be disproved). UsingNoda’s rule and a comparison of the actual FTIR spectra indicatesthe following order of diagenetic susceptibility (removal) at siteCM from shallowest to deepest: carboxyl groups (including ali-phatic ester and amide/protein) > carbohydrates > aromatic com-pounds and the SiAO framework in clay/biogenic silica. OAH/NAH groups change before aliphatic ester, amide/protein and thecarbohydrates indicated by 1092 cm�1. The SiAOH bond in bio-genic silica (3621 cm�1) also changes (decrease in relative concen-tration at CM) before the OAH/NAH bonds resulting frompolymeric alcohols/carboxylic acid/amide or absorbed water andcarbohydrates. This order of response is based upon data in Table 3and Fig. 7 (see Supplementary S-4 for further details).

Similar trends were observed for EM (Supplementary Figs. S-3and S-4) except that preferential changes in all carbohydrate peaksoccur before changes in OAH/NAH groups and SiAOH bond. Thisindicates that the activity of OAH/NAH bonds in polymeric alco-hol/carboxylic acid/amide/silicate bands/adsorbed water, as wellas SiAOH bonds in biogenic silica are site dependent; in otherwords, the compounds which contribute to the OAH/NAH bandsin each station are different, leading to the OAH/NAH activity dif-ferences. In addition, the diagenetic status of biogenic silica may bedifferent among stations.

Therefore, for CM and EM stations the general sequential orderof OM diagenesis in sediment is that carboxyl groups (includingaliphatic ester and amide/protein) are the first to exhibit losses,followed by a group of carbohydrates, while clay minerals/SiAOframework in biogenic silica and the aromatic compounds exhibitgreater stability and thus an increase in relative FTIR response. Theactivity of OAH/NAH bonds in polymeric alcohols/carboxylic acid/amide/adsorbed water bands as well as SiAOH bands in biogenicsilica are site dependent.

NM, SM and WM show a similar diagenetic tendency for all theother functional groups (with slight variation in peak position)compared with CM and EM, except for the clay/biogenic silica/car-bohydrate bands (as shown in Fig. 7b for WM and Supplementary,

Fig. 7. Asynchronous 2D correlation spectrum for (a) CM and (b) WM. Each spectral data set was constructed sorting from shallowest depth to deepest. The data set wasnormalized and PCA reconstructed before 2D correlation analysis. For ease in identification of peaks, horizontal slices (y is constant) across the region of interest are plottedbelow each 2D correlation spectrum.

134 H. Li et al. / Organic Geochemistry 58 (2013) 125–136

Fig. S-3b and c, for the other sites). At NM, SM and WM, the changein clay/biogenic silica/carbohydrate bands (1100–1000 cm�1) ap-pears to occur after all other functional groups. This could be dueto the intense decomposing of a thin layer of fresh OM at the sed-iment surface of these sites vs. CM and EM, and the dominance ofinactive clay or relatively stable OM moieties (perhaps terrestri-ally-derived structural carbohydrates). With the diagenetic lossof total OM with depth, the relative proportion of these refractorycomponents would increase, as also shown by synchronous spectrafor NM, SM and WM (see above and Fig. 6).

The compositional differences characterized by a domination ofinactive clays/biogenic silica/stable carbohydrate at the sites couldbe caused by the differences in sediment source, sedimentationrate, water column depth and depositional environment, such asintensity of sediment re-suspension. It is possible that the wind-driven prevailing circulation in the lake contributes more terres-trial material at certain times of the year to NM, carrying a rela-tively large amount of non-active clay minerals and refractorycarbohydrate. Adding to this likelihood is the relative closenessof NM to several large inflows, including the Nipigon River, PigeonRiver, Pic River and Aguasabon River systems as well the site’slocation relative to the Slate Islands. For SM, as discussed above,

the refractory nature of the sediment is most likely due to a localacceleration of bottom currents in the deep trough, inhibiting theaccumulation of fine-grained sediment, leaving coarser sedimentwith refractory, reworked OM. For WM, the clay/biogenic silica/carbohydrate response may result from the age of the surface sed-iments (Table 1); the WM sediment core exhibits an age range of940–7910 yr BP in the top 10 cm. These ages are significantlygreater than those at EM (235 and 1530 yr BP for the 0–2 and 8–10 cm slices, respectively). As WM surface-sediment OM is older,the labile carbohydrates may already have been degraded or trans-ported elsewhere.

The diagenesis trends are consistent with previous studies ofOM diagenesis in both marine and lacustrine sediments. For exam-ple, sediments from the southern Cape coastline of South Africaanalyzed using pyrolysis–gas chromatography mass spectrometryshow that selective preservation/degradation drives the majordown core variability in OM composition (Carr et al., 2010). Thelower half of the core, older than 12,000 yr, was characterized bysuites of low MW aromatic pyrolysis products; surface sedimentswere characterized by products derived from fresh emergent orterrestrial vegetation, including lignin monomers, plant-derivedfatty acids and long chain n-alkanes. A 13C NMR study of sediment

H. Li et al. / Organic Geochemistry 58 (2013) 125–136 135

diagenesis in Jellyfish Lake, an anoxic marine lake in Palau (Oremet al., 1991) revealed that the major post-deposition change in sed-imentary OM was carbohydrate biodegradation, whereas ligninand aliphatic substances were preserved.

4. Conclusions

Isotopic analysis, C and N content and FTIR coupled with PCAand 2D correlation analyses were used to study the compositionand diagenesis of Lake Superior sediments. Dominant signals fromclay minerals/biogenic silica along with peaks from common OMfunctional groups (carbohydrate, carboxylic acid, aliphatic/acetylester, amide/protein and phenol/lignin) are identified; their diage-netic changes are also studied.

The stable carbon isotopic values and C/N values at all sites(Fig. 2) are consistent with primarily microbial OM, either froman autochthonous or soil microbial source. The %C concentrationin the sediments ranged from just over 4% to ca. 1% and decreasedwith depth at all sites, showing the strongest degree of change atWM (and a high degree of change in SM as well, though we haveincomplete data for that site).

PCA analysis of the FTIR data (Fig. 5) shows that, with increasingdepth, carboxylates, aliphatic esters, amide/protein and carbohy-drates decrease and the remaining FTIR amenable material be-comes more similar across the lake sites. The greatest spreadalong PC 1 with depth occurs for site SM, followed by site WM,as would be consistent with the degree of change in %C in the sur-face sediments across sites (Fig. 2).

2D correlation analysis, by spreading FTIR spectra into a seconddimension, increased the resolution, separating overlapping bandsand providing more detailed information about band positions andchanges in band intensity with diagenesis. As with the PCA analy-sis, 2D correlation analysis (Fig. 6, Table 3) shows that the compo-sition of the sediments changes with depth. Aliphatic esterfunctional group and some carbohydrates appear most susceptibleto diagenesis. Slight shifts in wavenumber indicate changes withinthese compound classes as a function of location in the lake. Asyn-chronous spectra (Fig. 7) can provide a sequential order of earlystage OM diagenesis. In general: carboxyl groups including thosein aliphatic ester and amide/protein change first, followed by agroup of carbohydrates and then aromatic compounds and/or clayminerals and/or the SiAO framework of biogenic silica. This is con-sistent with bulk information showing initial loss of nitrogen rela-tive to carbon (which could be due to rapid diagenetic changes inprotein concentration).

In addition to general diagenetic trends in Lake Superior sedi-ment, we do see some spatial variability in surface sediment com-position. NM, SM and WM have distinct down core compositionalvariation relative to EM and CM. We believe NM to be affected byincreased terrestrial input, SM by local acceleration of bottom cur-rents in the deep trough and WM by the geological nature of thesediment or sediment resuspension. In addition, down-core vari-ability in the lake may be related to variation in sedimentationrate. Future research coupling FTIR analysis of deeper cores alongwith estimated sedimentation rates for each core would be a usefulapproach for further study.

Acknowledgements

We thank the captain and crew of the R/V Blue Heron for assis-tance with sample collection as well as H. Abdulla for comments onearly versions of the manuscript. J. Jacob and four anonymousreviewers are also thanked for the comments that greatly im-proved the final version. The research was supported by NSF GrantOCE-0825600 (to E.C.M.).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.orggeochem.2013.03.002.

Associate Editor—P.A. Meyers

References

Abdulla, H.A.N., Minor, E.C., Dias, R.F., Hatcher, P.G., 2010a. Changes in thecompound classes of dissolved organic matter along an estuarine transect: astudy using FTIR and 13C NMR. Geochimica et Cosmochimica Acta 74, 3815–3838.

Abdulla, H.A.N., Minor, E.C., Hatcher, P.G., 2010b. Using two-dimensionalcorrelations of 13C NMR and FTIR to investigate changes in the chemicalcomposition of dissolved organic matter along an estuarine transect.Environmental Science & Technology 44, 8044–8049.

Assel, R.A., 1986. Fall and winter thermal structure of Lake Superior. Journal of GreatLakes Research 12, 251–262.

Baker, J.E., Eisenreich, S.J., Eadie, B.J., 1991. Sediment trap fluxes and benthicrecycling of organic carbon, polycyclic aromatic hydrocarbons, andpolychlorobiphenyl congeners in Lake Superior. Environmental Science &Technology 25, 500–509.

Ball, D.E., 1964. Loss-on-ignition as an estimate of organic matter and organiccarbon in non-calcareous soils. Journal of Soil Science 15, 84–92.

Ben-Dor, E., Banin, A., 1989. Determination of organic matter content in arid-zonesoils using a simple ‘‘loss-on-ignition’’ method. Communicationsin Soil Science and Plant Analysis 20, 1675–1695.

Bennington, V., McKinley, G.A., Kimura, N., Wu, C.H., 2010. General circulation ofLake Superior: mean, variability, and trends from 1979 to 2006. Journal ofGeophysical Research 115, C12015. http://dx.doi.org/10.1029/2010JC006261.

Bruland, K.W., Koide, M., Bowser, C., Maher, L.J., Goldberg, E.D., 1975. 210Pb andpollen geochronologies on Lake Superior sediments. Quaternary Research 5, 89–98.

Carr, A.S., Boom, A., Chase, B.M., Roberts, D.L., Roberts, Z.E., 2010. Molecularfingerprinting of wetland organic matter using pyrolysis-GC/MS: an examplefrom the southern Cape coastline of South Africa. Journal of Paleolimnology 44,947–961.

Celi, L., Schnitzer, M., Nègre, M., 1997. Analysis of carboxyl groups in soil humicacids by a wet chemical method, Fourier transform infrared spectrometry andsolution-state 13C nuclear magnetic resonance, a comparative study. SoilScience 162, 189–197.

Degens, E.T., 1967. Diagenesis of organic matter. In: Larsen, G., Chilingar, G.V. (Eds.),Diagenesis in Sediments. Elsevier, Amsterdam, pp. 345–390.

Derkacheva, O., Sukhov, D., 2008. Investigation of lignins by FTIR spectroscopy.Macromolecular Symposia 265, 61–68.

Dutka, B.J., Bell, J.B., Liu, D.L.S., 1974. Microbiological examination of offshore LakeErie sediments. Journal of the Fisheries Research Board of Canada 31, 299–308.

Evans, J.E., Johnson, T.C., Alexander Jr., E.C., Lively, R.S., Eisenreich, S.J., 1981.Sedimentation rates and depositional processes in Lake Superior from 210Pbgeochronology. Journal of Great Lakes Research 7, 299–310.

Griffiths, P.R., De Haseth, J.A., 2007. Fourier Transform Infrared Spectroscopy. Wiley-Interscience, New Jersey.

Harisa, P.I., Severcanb, F., 1999. FTIR spectroscopic characterization of proteinstructure in aqueous and non-aqueous media. Journal of Molecular Catalysis B:Enzymatic 7, 207–221.

Harris, D., Horwáth, W.R., Van Kessel, C., 2001. Acid fumigation of soils to removecarbonates prior to total organic carbon or carbon-13 isotopic analysis. SoilScience Society of America Journal 65, 1853–1856.

Herbert, T.D., Tom, B.A., Burnett, C., 1992. Precise major component determinationsin deep-sea sediments using Fourier transform infrared spectroscopy.Geochimica et Cosmochimica Acta 56, 1759–1763.

Herdendorf, C.E., 1990. Distribution of the world’s large lakes. In: Tilzer, M.M.,Serruya, C. (Eds.), Large Lakes: Ecological Structure and Function. Springer-Verlag, NY, pp. 3–38.

Hoffman, J.C., Peterson, G.S., Cotter, A.M., Kelly, J.R., 2010. Using stable isotopemixing in a Great Lakes coastal tributary to determine food web linkages inyoung fishes. Estuaries and Coasts. http://dx.doi.org/10.1007/s12237-010-9295-0.

Johnson, T.C., 1980. Late-glacial and postglacial sedimentation in Lake Superiorbased on seismic-reflection profiles. Quaternary Research 13, 380–391.

Johnson, T.C., Eisenreich, S.J., 1979. Silica in Lake Superior: mass balanceconsiderations and a model for dynamic response to eutrophication.Geochimica et Cosmochimica Acta 43, 7–92.

Johnson, T.C., Evans, J.E., Eisenreich, S.J., 1982. Total organic carbon in Lake SuperiorUSA, Canada: sediments comparisons with hemi pelagic and pelagic marineenvironments. Limnology and Oceanography 27, 481–491.

Jung, Y.M., 2004. Principal component analysis based two-dimensional correlationspectroscopy for noise filtering effect. Vibrational Spectroscopy 36, 267–270.

Kemp, A.L.W., Johnston, L.M., 1979. Diagenesis of organic matter in the sediments ofLakes Ontario, Erie, and Huron. Journal of Great Lakes Research 5, 1–10.

136 H. Li et al. / Organic Geochemistry 58 (2013) 125–136

Klump, J.V., Paddock, R., Remsen, C.C., Fitzgerald, S., Boraas, M., Anderson, P., 1989.Variations in sediment accumulation rates and the flux of labile organic matterin eastern Lake Superior basins. Journal of Great Lakes Research 15, 104–122.

Krezoski, J.R., 1989. Sediment reworking and transport in eastern Lake Superior:in situ rare earth element tracer studies. Journal of Great Lakes Research 15, 26–33.

Kvalheim, O.M., Aksnes, D.W., Brekke, T., Eide, M.O., Sletten, E., 1985. Crude oilcharacterization and correlation by principal component analysis of 13C nuclearmagnetic resonance spectra. Analytical Chemistry 57, 2858–2864.

Lehmann, M.F., Bernasconi, S.M., Barbieri, A., McKenzie, J.A., 2002. Preservation oforganic matter and alteration of its carbon and nitrogen isotope compositionduring simulated and in situ early sedimentary diagenesis. Geochimica etCosmochimica Acta 66, 3573–3584.

Li, J., Miklesh, D., Kistner, M., Canfield, D.E., Katsev, S., 2012. Carbon mineralizationand oxygen dynamics in sediments with deep oxygen penetration, LakeSuperior. Limnology & Oceanography 57, 1634–1650.

MacCarthy, P., Rice, J.A., 1985. Spectroscopic methods (other than NMR) fordetermining functionality. In: Aiken, G., MCknight, D., Wershaw, R., MacCarthy,P. (Eds.), Humic Substances in Soil, Sediment, and Water: Geochemistry,Isolation and Characterization. John Wiley and Sons, New York, NY, pp. 527–559.

Madejová, J., 2003. FTIR techniques in clay mineral studies. VibrationalSpectroscopy 31, 1–10.

Madejová, J., Komadel, P., 2001. Baseline studies of the clay minerals society sourceclays: infrared methods. Clays and Clay Minerals 49, 410–432.

Mascarenhas, J.D., Arbuckle, G.A., 2000. Characterization of plant carbohydrates andchanges in leaf carbohydrate chemistry due to chemical and enzymaticdegradation measured by microscopic ATR FTIR spectroscopy. Society forApplied Spectroscopy 54, 681–686.

Matheson, D.H., Munawar, M., 1978. Lake Superior basin and its development.Journal of Great Lakes Research 4, 249–263.

McManus, J., Heinen, E.A., Baehr, M.M., 2003. Hypolimnetic oxidation rates in LakeSuperior: role of dissolved organic material on the lake’s carbon budget.Limnology and Oceanography 48, 1624–1632.

Mecozzi, M., Pietroletti, M., Gallo, V., Conti, M.E., 2009a. Formation of incubatedmarine mucilages investigated by FTIR and spectroscopy and supported by two-dimensional correlation analysis. Marine Chemistry 116, 18–35.

Mecozzi, M., Pietrantonio, E., Pietroletti, M., 2009b. The roles of carbohydrates,proteins and lipids in the process of aggregation of natural marine organicmatter investigated by means of 2D correlation spectroscopy applied to infraredspectra. Spectrochimica Acta Part A 71, 1877–1884.

Mecozzi, M., Moscato, F., Pietroletti, M., Quarto, F., Oteri, F., Cicero, A.M.,2009c. Applications of FTIR spectroscopy in environmental studiessupported by two dimensional correlation analysis. Global NEST Journal11, 593–600.

Meyers, P.A., Ishiwatari, R., 1993. Lacustrine organic geochemistry – an overview ofindicators of organic matter sources and diagenesis in lake sediments. OrganicGeochemistry 20, 867–900.

Noda, I., 1993. Generalized two-dimensional correlation method applied to infrared,Raman, and other types of spectroscopy. Society for Applied Spectroscopy 47,1329–1336.

Noda, I., Ozaki, Y., 2004. Two-Dimensional Correlation Spectroscopy: Applicationsin Vibrational and Optical Spectroscopy. John Wiley and Sons Inc., SanFrancisco, USA.

Orem, W.H., Burnett, W.C., Landing, W.M., Lyons, W.B., Shower, W., 1991. Jellyfishlake, Palau: early diagenesis of organic matter in sediments of an anoxic marinelake. Limnology and Oceanography 36, 526–543.

Ostrom, N.E., Long, D.T., Bell, E.M., Beals, T., 1998. The origin and cycling ofparticulate and sedimentary organic matter and nitrate in Lake Superior.Chemical Geology 152, 13–28.

Pandurangi, R.S., Seehra, M.S., Razzaboni, B.L., Bolsaitist, P., 1990. Surface and bulkinfrared modes of crystalline and amorphous silica particles: a study of therelation of surface structure to cytotoxicity of respirable silica. EnvironmentalHealth Perspectives 86, 327–336.

Popescu, C., Popescu, M., Vasile, C., 2010. Characterization of fungal degraded limewood by FTIR and 2D IR correlation spectroscopy. Microchemical Journal 95,377–387.

Raiskila, S., Pulkkinen, M., Laakso, T., Fagerstedt, K., Löija, M., Mahlberg, R., Paajanen,L., Ritschkoff, A.-C., Saranpää, P., 2007. FTIR spectroscopic prediction of klasonand acid soluble lignin variation in Norway spruce cutting clones. Silva Fennica41, 351–371.

Ramachandran, V.S., Beaudoin, J.J., 2008. Handbook of Analytical Techniques inConcrete Science and Technology: Principles, Techniques and Applications.Noyes Publications/William Andrew Publishing, LLC., Norwich, New York, USA.

Sanni, O.D., Wagner, M.S., Briggs, D., Castner, D.G., Vickerman, J.C., 2002.Classification of adsorbed protein static ToF-SIMS spectra by principalcomponent analysis and neural networks. Surface and Interface Analysis 33,715–728.

Silverstein, R.M., Webster, F.X., 1997. Spectrometric Identification of OrganicCompounds. John Wiley & Sons, Inc., New York.

Smith, B.C., 1996. Fundamentals of Fourier Transform Infrared Spectroscopy. CRCPress, Boca Raton, Florida.

Smith, B.C., 1999. Infrared Spectral Interpretation: A Systematic Approach. CRCPress, Boca Raton, Florida.

Stephens, B.M., Minor, E.C., 2010. DOM characteristics along the continuum fromriver to receiving basin: a comparison of freshwater and marine transects.Aquatic Sciences 72, 403–417.

Stuiver, M., Polach, H.A., 1977. Discussion: reporting of 14C data. Radiocarbon 19,355–363.

Urban, N.R., Lu, X., Chai, Y., Apul, D.S., 2004. Sediment trap studies in Lake Superior:insights into resuspension, cross-margin transport, and carbon cycling. Journalof Great Lakes Research 30, 147–161.

Urban, N.R., Auer, M.T., Green, S.A., Lu, X., Apul, D.S., Powell, K.D., Bub, L., 2005.Carbon cycling in Lake Superior. Journal of Geophysical Research 110, C06S90.http://dx.doi.org/10.1029/2003JC002230.

Vanderpost, J.M., 1972. Bacterial and physical characteristics of Lake Ontariosediment during several months. In: Proceedings 15th Conference of GreatLakes Research, pp. 198–213.

Vogel, H., Rosen, P., Wagner, B., Melles, M., Persson, P., 2008. Fourier transforminfrared spectroscopy, a new cost-effective tool for quantitative analysis ofbiogeochemical properties in long sediment records. Journal of Paleolimnology40, 698–702.

Widjanarko, S.B., Nugroho, A., Estiasih, T., 2011. Functional interaction componentsof protein isolates and glucomannan in food bars by FTIR and SEM studies.African Journal of Food Science 5, 12–21.

Zigah, P.K., Minor, E.C., Werne, J.P., 2012. Radiocarbon and stable-isotopegeochemistry of organic and inorganic carbon in Lake Superior. GlobalBiogeochemical Cycles 26, GB1023. http://dx.doi.org/10.1029/2011GB004132.