influence of bedrock geology on dissolved organic matter...

12
Influence of bedrock geology on dissolved organic matter quality in stream water Jennifer J. Mosher a,1 , Geoffrey C. Klein b,c,2 , Alan G. Marshall b,c , Robert H. Findlay a,a Department of Biological Sciences, University of Alabama, Tuscaloosa, AL 35487, United States b National High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32306-4005, United States c Department of Chemistry and Biochemistry, Florida State University, Tallahassee, FL 32306-4390, United States article info Article history: Received 14 July 2009 Received in revised form 20 July 2010 Accepted 14 August 2010 Available online 20 August 2010 abstract A field study was conducted to determine the linkage between dissolved organic matter (DOM) quality and bedrock geology. Sediment and water samples were collected from six streams, three flowing over sandstone and three flowing over limestone bedrock in the Bankhead National Forest, Alabama. Various stream water and sediment characteristics were determined by standard limnological methods and DOM quality was determined by Fourier transform ion cyclotron resonance mass spectrometry with sample introduction by electrospray ionization (ESI FT-ICR MS). Stream waters from limestone streams showed higher concentrations of calcium, magnesium, sodium, sulfate and dissolved inorganic carbon, as well as higher pH and acid neutralizing capacity, and were enriched in condensed hydrocarbons. In contrast, stream waters from sandstone streams showed higher concentrations of aluminum and iron and were enriched in tannin-like compounds. ESI FT-ICR MS analysis indicated that stream water DOM consisted of between 2950 and 3700 individual compounds ranging from 250 to 1000 Dalton. DOM compounds were grouped into 35 heteroatom (N n O o S s ) classes based on molecular formula assignments and multi- variate statistical analysis was applied to reveal the patterns of variation of the DOM quality and identify environmental parameters associated with the variation. Partial redundancy analysis indicated that the variation in stream water DOM quality correlated with stream location and water chemistry parameters associated with bedrock type. A similar analysis including three previously published ESI FT-ICR MS anal- yses of stream water DOM showed similar correlations with location and stream water chemistry, indi- cating that bedrock geology via stream water chemistry influences DOM quality. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction Dissolved organic matter (DOM) in stream ecosystems is an essential component for many geochemical processes and is a ma- jor energy source for microbially based food webs. Stream water DOM is a complex mixture of thousands of organic molecules vary- ing in size, molecular composition and biological and chemical lability. This complex mixture results, in part, because stream water DOM originates from a combination of aquatic and terrige- nous sources, each of which are themselves a source of complex DOM. In forested streams, DOM is primarily derived from terrige- nous organic matter, which reaches the stream via overland flow or groundwater, with a large proportion originating from the forest floor and/or organic horizon of the soil (Kaplan and Newbold, 1993; Aitkenhead-Peterson et al., 2003), although algal production and heterotrophic decomposition within the stream can contribute autochthonous DOM. This complexity has hindered the study of the central role of DOM in stream ecosystems. The number, nature and relative abundance of molecules that comprise stream water DOM is often referred to as DOM quality, as these characteristics affect its many roles in stream ecosystems (Kim et al., 2006). Several landscape level factors influence DOM quality in stream water. Biotic factors include terrigenous flora composition, autochthonous production and microbial decomposi- tion, both in stream and in the surrounding terrigenous watershed (Meier et al., 2004; Schumacher et al., 2006). Physical and chemical factors include hydrology, depth to water table, watershed slope, photooxidation, precipitation and soil moisture (Maurice and Leff, 2002; Frost et al., 2006). Minerals are known to influence DOM quality in stream water through sorption in soils: this process re- moves or decreases the relative abundance of particular chemical classes during their production and transport from the terrigenous environment to the stream (Maurice et al., 2002). The characterization of DOM quality has evolved from deter- mining generalized categories of size and lability to high resolution techniques capable of providing molecular level identification. Of 0146-6380/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.orggeochem.2010.08.004 Corresponding author. Address: 2105 Bevill Building, Box 870206, Department of Biological Sciences, University of Alabama, Tuscaloosa, AL 35487, United States. Tel.: +1 205 348 4167; fax: +1 205 348 1403. E-mail address: rfi[email protected] (R.H. Findlay). 1 Present address: Biosciences Division, Oak Ridge National Laboratory, Oak Ridge, TN 37831, United States. 2 Present address: Department of Biology, Chemistry and Environmental Science, Christopher Newport University, Newport News, VA 23606, United States. Organic Geochemistry 41 (2010) 1177–1188 Contents lists available at ScienceDirect Organic Geochemistry journal homepage: www.elsevier.com/locate/orggeochem

Upload: others

Post on 14-Aug-2020

5 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

Organic Geochemistry 41 (2010) 1177–1188

Contents lists available at ScienceDirect

Organic Geochemistry

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

Influence of bedrock geology on dissolved organic matter quality in stream water

Jennifer J. Mosher a,1, Geoffrey C. Klein b,c,2, Alan G. Marshall b,c, Robert H. Findlay a,⇑a Department of Biological Sciences, University of Alabama, Tuscaloosa, AL 35487, United Statesb National High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32306-4005, United Statesc Department of Chemistry and Biochemistry, Florida State University, Tallahassee, FL 32306-4390, United States

a r t i c l e i n f o

Article history:Received 14 July 2009Received in revised form 20 July 2010Accepted 14 August 2010Available online 20 August 2010

0146-6380/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.orggeochem.2010.08.004

⇑ Corresponding author. Address: 2105 Bevill Buildof Biological Sciences, University of Alabama, TuscaloTel.: +1 205 348 4167; fax: +1 205 348 1403.

E-mail address: [email protected] (R.H. Findla1 Present address: Biosciences Division, Oak Ridge N

TN 37831, United States.2 Present address: Department of Biology, Chemistr

Christopher Newport University, Newport News, VA 23

a b s t r a c t

A field study was conducted to determine the linkage between dissolved organic matter (DOM) qualityand bedrock geology. Sediment and water samples were collected from six streams, three flowing oversandstone and three flowing over limestone bedrock in the Bankhead National Forest, Alabama. Variousstream water and sediment characteristics were determined by standard limnological methods and DOMquality was determined by Fourier transform ion cyclotron resonance mass spectrometry with sampleintroduction by electrospray ionization (ESI FT-ICR MS). Stream waters from limestone streams showedhigher concentrations of calcium, magnesium, sodium, sulfate and dissolved inorganic carbon, as well ashigher pH and acid neutralizing capacity, and were enriched in condensed hydrocarbons. In contrast,stream waters from sandstone streams showed higher concentrations of aluminum and iron and wereenriched in tannin-like compounds. ESI FT-ICR MS analysis indicated that stream water DOM consistedof between 2950 and 3700 individual compounds ranging from 250 to 1000 Dalton. DOM compoundswere grouped into 35 heteroatom (NnOoSs) classes based on molecular formula assignments and multi-variate statistical analysis was applied to reveal the patterns of variation of the DOM quality and identifyenvironmental parameters associated with the variation. Partial redundancy analysis indicated that thevariation in stream water DOM quality correlated with stream location and water chemistry parametersassociated with bedrock type. A similar analysis including three previously published ESI FT-ICR MS anal-yses of stream water DOM showed similar correlations with location and stream water chemistry, indi-cating that bedrock geology via stream water chemistry influences DOM quality.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Dissolved organic matter (DOM) in stream ecosystems is anessential component for many geochemical processes and is a ma-jor energy source for microbially based food webs. Stream waterDOM is a complex mixture of thousands of organic molecules vary-ing in size, molecular composition and biological and chemicallability. This complex mixture results, in part, because streamwater DOM originates from a combination of aquatic and terrige-nous sources, each of which are themselves a source of complexDOM. In forested streams, DOM is primarily derived from terrige-nous organic matter, which reaches the stream via overland flowor groundwater, with a large proportion originating from the forestfloor and/or organic horizon of the soil (Kaplan and Newbold,

ll rights reserved.

ing, Box 870206, Departmentosa, AL 35487, United States.

y).ational Laboratory, Oak Ridge,

y and Environmental Science,606, United States.

1993; Aitkenhead-Peterson et al., 2003), although algal productionand heterotrophic decomposition within the stream can contributeautochthonous DOM. This complexity has hindered the study ofthe central role of DOM in stream ecosystems.

The number, nature and relative abundance of molecules thatcomprise stream water DOM is often referred to as DOM quality,as these characteristics affect its many roles in stream ecosystems(Kim et al., 2006). Several landscape level factors influence DOMquality in stream water. Biotic factors include terrigenous floracomposition, autochthonous production and microbial decomposi-tion, both in stream and in the surrounding terrigenous watershed(Meier et al., 2004; Schumacher et al., 2006). Physical and chemicalfactors include hydrology, depth to water table, watershed slope,photooxidation, precipitation and soil moisture (Maurice and Leff,2002; Frost et al., 2006). Minerals are known to influence DOMquality in stream water through sorption in soils: this process re-moves or decreases the relative abundance of particular chemicalclasses during their production and transport from the terrigenousenvironment to the stream (Maurice et al., 2002).

The characterization of DOM quality has evolved from deter-mining generalized categories of size and lability to high resolutiontechniques capable of providing molecular level identification. Of

Page 2: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

1178 J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188

these techniques, Fourier transform ion cyclotron resonance massspectrometry (Marshall et al., 1998) coupled with sample introduc-tion by electrospray ionization (ESI FT-ICR MS) has shown thegreatest promise in the analysis of DOM quality in natural waters(Kujawinski et al., 2002; Kim et al., 2003a). This technique resolvescompounds of different elemental composition (chemical formula,CcHhNnSsOoPp) and determines their relative abundance. Themolecular mass of each compound is determined with sufficientaccuracy (mass errors <1 ppm) that unique chemical formulascan be assigned, although the relative abundances that are deter-mined suffer from an unresolved problem of differing ionizationefficiencies among chemical classes (Kim et al., 2006).

Studies utilizing ESI FT-ICR MS to examine DOM quality inaquatic ecosystems are relatively rare. Recent methodologicalrefinements allowing the determination of the molecular formulasof more than 4000 individual molecules provided, for the first time,the ability to determine molecular level measurements of DOMquality (Kujawinski et al., 2002; Stenson et al., 2003). Comparisonsbetween samples were made by visual observation of Kendrickmass defect (KMD) versus Kendrick nominal mass plots. Theseplots allow for visualization of classes of carbon distribution (com-positions differing by multiples of –CH2) (Hughey et al., 2001). Kimet al. (2003a) introduced the use of van Krevelen diagrams for a vi-sual representation of possible reaction pathways and for qualita-tive analysis of major classes of compounds. However, theseapproaches do not allow direct statistical analysis of DOM quality:to date, only one such study has been reported (Koch et al., 2005).The authors compared the DOM quality of seawater from WeddellSea, Antarctica and pore water from a tropical mangrove forest byBray-Curtis similarity analysis. They found that the majority ofDOM compounds were present in both types of samples.

To determine the nature and extent that bedrock mineral com-position influences stream water DOM quality, we sampled waterand sediments from six streams, three flowing over sandstone bed-rock and three flowing over limestone bedrock (subsequently re-ferred to as sandstone and limestone streams). We measuredcation and anion concentrations along with other stream waterand sediment characteristics by standard limnological methodsand used ESI FT-ICR MS to determine stream water DOM quality.We utilized multivariate statistical analysis to compare DOM qual-ity among the sandstone and limestone streams and to determinecorrelations among stream water chemistry, total sediment micro-bial biomass, sediment chlorophyll a, sediment grain size and vari-ations in DOM quality. We extended this study by incorporatingpreviously published DOM quality data (Kim et al., 2003b, 2006)from streams located in three distant watersheds to determine ifpatterns in the influence of bedrock type on DOM quality couldbe extended beyond the geographic boundaries of our study area.This comparison is subsequently referred to as the among wa-tershed comparison.

2. Methods

2.1. Site description

Six streams, three flowing over limestone bedrock and threeflowing over sandstone bedrock were studied (Fig. 1). The studyarea was in the William B. Bankhead National Forest, in northernAlabama, USA directly south of the Tennessee Valley Divide. Allstreams are second or third order tributaries to the Sipsey Forkof the Black Warrior River system in the Mobile River drainage ba-sin. All streams have similar elevations, discharge, sediment grainsize and are located within 21 km of each other in the north centralportion of the forest. The creek beds consist of exposed bedrock,cobble and unconsolidated sediment.

The forest is located in the Appalachian Plateau physiographicprovince and lies within the Plateau Coal Region of the WarriorCoal Field. The Plateau Coal Region is characterized by discontinu-ous sulfur rich bituminous coal bearing plateaus separated bydownsloping valleys (Ward and Chase, 1981). However, due tothe Sipsey Wilderness Area (a highly protected region within theNational Forest), the amount and location of many of the coal bear-ing plateaus has not been quantified and required the use of partialredundancy analysis (see Section 2.7) to account for this variabil-ity. The primary bedrock within the region is Pennsylvanian agePottsville Formation comprising of sandstone/shale with thin lay-ers of discontinuous coal (Diehl et al., 2004). Some valleys andstreambeds in this region have eroded through the Pottsville bed-rock to older Mississippian age bedrock, specifically Parkwoodsandstone and Bangor limestone formations (Fig. 1). Two of thestreams (Beech and Brushy creeks) flow over exposed ParkwoodFormation bedrock, characterized by shale/sandstone and clayeycoal. Three streams (Thompson, Flannagin and Borden creeks) haveeroded through both the Pottsville and Parkwood bedrock and flowover exposed Bangor limestone. Only one stream in this study(Hubbard Creek) flows solely over the overlying Pottsville Forma-tion sandstone. The streams that have eroded through the upperlayer(s) still contain Pottsville Formation sandstone in the uplandportion of their watershed (Mast and Turk, 1999). Stream flowwas from generally north to south; streams were located withinthe forest (moving from west to east) as Hubbard [sandstone],Thompson [limestone], Flannagin [limestone], Borden [limestone],Beech [sandstone] and Brushy [sandstone] with Hubbard andThompson, Flannagin and Borden, and Beech and Brushy formingproximal pairs. The catchments are heavily forested, primarily bymixed hardwoods interspersed with pine stands and the cool,moist conditions of the stream valleys supported Appalachian flora(Mast and Turk, 1999). We noted that Tsuga canadensis (EasternHemlock) was present only in the sandstone stream riparian zones.

2.2. Sampling

Sampling was conducted in March 2006 during base flow condi-tions, two weeks following the last recorded precipitation in thearea. Sediment from each stream was sampled in triplicate with apush core. The top 1 cm of sediment of each core was homogenizedand subsamples taken for each measurement. Stream water was col-lected in duplicate and physical parameters were measured in situwith portable meters. All samples were placed on ice and either pro-cessed immediately upon return to the laboratory or preserved (fro-zen and/or acidified) and stored until analysis.

2.3. Physical parameters

Total suspended solids (TSS) were determined by filtering wellmixed stream water samples through dried, preweighed glass fiberfilters (GF/F) (Whatman Inc., Florham Park, New Jersey) and dryingat 103 �C to a constant mass (Fanson, 1985). Specific surface area ofcombusted sediments was determined by the adsorption isothermof nitrogen under vacuum (Brunauer et al., 1938). Average weeklydischarge in the streams was estimated by a rating curve con-structed from instantaneous measurements of discharge and theplacement of barologgers in each stream (Levelogger 3001, SolinstCanada Ltd., Georgetown, Ontario, Canada). The rating curve wasconstructed from monthly (n = 15) measurements of stream veloc-ity (Marsh McBirney, Frederick, Maryland), depth measurementsand width. The average weekly gauge height obtained from thebarologgers was applied to the rating curve, and discharge forthe week preceding the date of sampling was estimated. Canopycover was estimated with a handheld convex spherical crown den-sitometer (Forestry Suppliers, Jackson, Mississippi). Stream loca-

Page 3: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

Hubbard

Thompson

Flannagin

BrushyBeech

Borden

Fig. 1. Map of the study site showing bedrock type. Light blue represents Pottsville Formation, purple the Parkwood Formation, and dark blue Bangor Limestone. The blackcircles indicate the sampling sites, the vertical line emphasizes the western forest boundary and the horizontal line with crosshatches at latitude 34.1225 �N illustrates howthe covariable location was determined for the six-stream study. The image was scanned from Geological Survey of Alabama Special Map 220 (Mancini, 1988).

J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188 1179

tion was determined by measuring the distance of each streamfrom the western edge of the forest at a latitude of 33�06.5540 byuse of Alabama Geological Survey special map 220 (Szabo andCopeland, 1988). Catchment area and stream slope were calculatedfrom United State Geological Survey 7.5 min series topographicmaps. Stream location was used as a synthetic descriptor variable(Magalhaes et al., 2002) to represent additional factors potentiallyinfluencing DOM quality (e.g., forest composition, catchment coalcontent) but not measured in this study.

2.4. Chemical parameters

Stream water samples for nutrients and dissolved organic car-bon (DOC) concentrations were taken with high density polyethyl-ene bottles (HDPE) bottles rinsed with sulfuric acid/nochromix�

mixture and washed with phosphate-free soap. The first 20 mL ofsample was filtered over a combusted GF/F, acidified with twodrops of 2 N optima grade HCl, stored at 4 �C until analysis andDOC concentrations analyzed by flash combustion in a ShimadzuTotal Organic Carbon Analyzer TOC-500. The remainder of thestream water was filtered through the same GF/F, stored at �20�C and analyzed for NO3–N, NO2–N, NH4–N and PO4–P concentra-tions in a Lachat QuickChem 8000.

Water samples for anions (Cl� and SO2�4 ), cations (Al3+, Ba2+,

Fe2+,3+, Na+, Ca2+, Mg2+, Mn2+,3+,4+, Si2+,4+ and K+) and acid neutralizingcapacity (ANC) were collected in HDPE bottles previously washedwith ultrapure water (>18.0 MX) and filtered over a combustedGF/F. Anion samples were stored at�20 �C and cation samples wereacidified (2% v:v) with optima grade nitric acid until analysis. Sam-ples for anion concentrations were analyzed by ion chromatographyand the cation concentration samples were subjected to inductivelycoupled plasma mass spectrometry (Welch et al., 1996).

ANC of the stream water was estimated by the Gran titrationmethod; pH was determined after incremental (0.025 mL) addi-tions of 0.1 N HCl to 60 mL stream water while stirring until fullprotonation had occurred. The Gran function was calculated fromEq. (1) (Wetzel and Likens, 2000).

F1 ¼ ðVorig þ VÞ½Hþ� ð1Þwhere F1 is the first Gran equation, Vorig is the volume of the sample, Vis the volume acid added and [H+] is the concentration of the hydro-gen ion. The equivalence point for the first Gran equation was foundby plotting F1 vs. acid volume and ANC was the value of F1 in meq/l atthe equivalence point. Dissolved inorganic carbon (DIC) was calcu-lated from ANC by use of Eq. (2) (Wetzel and Likens, 2000):

DIC ¼ ½½Alk� � ½OH�� þ ½Hþ��=½a1 þ 2a2� ð2Þ

where [Alk] is equivalent to ANC, [OH�1] is the concentration of thehydroxy ion and a1 and a2 are the decimal percentage that the car-bonate and bicarbonate ions comprise of total DIC.

Conductivity (Fisherbrand Traceable Digital 09-328, Fisher Sci-entific, Waltham, Massachusetts), pH (Oakton Instruments modelDouble-junction pH Testr, Vernon Hills, Illinois), dissolved oxygen(DO) and water temperature (YSI Model 95, Yellow Springs, Ohio)were measured in situ with hand held meters.

2.5. Biological parameters

Microbial biomass was estimated by phospholipid phosphateanalysis (Findlay et al., 1989). Cellular lipids were extracted fromthe sediments by dichloromethane/methanol/water extraction. To-tal microbial biomass was determined by the oxidation of lipidsand quantifying orthophosphate colorimetrically. A fraction of to-tal lipids, protected from light, was used to determine chlorophylla colorimetrically. Dried samples were suspended in 90% aqueousacetone and absorbance or optical density (OD) measured at 663,645 and 630 nm. Chlorophyll a abundance was calculated fromEq. (3) (Smoot et al., 1998):

Chlorophyll a ¼ 11:85ðOD663Þ � 1:54ðOD645Þ � 0:08ðOD630Þ ð3Þ

2.6. DOM quality

Stream water was collected in amber glass four liter bottles thathad been washed in phosphate-free soap, rinsed with ultrapure

Page 4: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

1180 J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188

water (>18.0 MX) and combusted at 450 �C for 4 h. One sample wastaken per stream with the exception of Thompson Creek for whichtriplicate samples were taken. Each sample was filtered, first witha combusted 0.7 lm GF/F then over a DI rinsed 0.2–0.3 lm filter(Millipore Corporation, Bedford, Massachusetts), and the pH ad-justed to 2.3–2.5 with optima grade HCl. DOM was extracted witha methanol cleaned C18 octadecyl disk (3M Empore, St. Paul, Minne-sota). Two 10 mL aliquots of 9:1 methanol:water were used to elutethe DOM into a clean flask under vacuum. The samples were concen-trated to 1 mL under a stream of nitrogen. Previous work by Kim et al.(2003b) demonstrated a 60% recovery of DOM with this techniquefrom stream water samples taken from the New Jersey Pine Barrens.

Ammonium hydroxide (5 ll, 30%, optima grade) was added to500 ll of each DOM sample to facilitate deprotonation. Sampleswere introduced into the mass spectrometer via a syringe pumpat 400 nl/min and a 50 lm i.d. fused silica micro ESI needle un-der typical ESI conditions (�2.0 kV; tube lens, �300 V; andheated capillary current, 2.4 A). Mass analysis was performedwith a custom built FT-ICR mass spectrometer equipped with a22 cm diameter horizontal bore 9.4 T actively shielded magnet(Oxford Corp., Oxney Mead, United Kingdom) (Senko et al.,1996a). Data were collected and processed with a modular ICRdata acquisition system (MIDAS) (Senko et al., 1996b). Ions wereaccumulated external to the magnet (Senko et al., 1997) in a lin-ear octopole ion trap (25.1 cm long) equipped with axial electricfield (Wilcox et al., 2002) for 20 s and transferred through rf-only multipoles to a 10 cm diameter, 30 cm long open cylindricalPenning ion trap. Multipoles were operated at an average of1.8 MHz at a peak-to-peak rf amplitude of 70 V. Broadband fre-quency sweep (‘‘chirp”) dipolar excitation (70 kHz to 1.27 MHzat a sweep rate of 150 Hz/ls and a peak-to-peak amplitude of190 V) was followed by direct mode image current detection(digitization rate at twice the highest excited spectral frequency,in this case 1.27 MHz) for 1.6 s to yield 4 Mword time-domaindata. The time-domain data were processed and Hanning-apod-ized, followed by a single zerofill before fast Fourier transforma-tion and magnitude calculation (Marshall and Verdun, 1990).Frequency was converted to mass-to-charge ratio (m/z) by thequadrupolar electric trapping potential approximation to gener-ate an m/z spectrum (Ledford et al., 1984; Shi et al., 2000).External mass calibration for negative ion ESI FT-ICR MS wasperformed with Agilent G2421A electrospray ‘‘tuning mix” (highmass) and stearic acid (low mass) as previously reported by Qianet al. (2001). This analysis resulted in a mass spectrum contain-ing 2950–3700 peaks per sample. Mass spectral peak height is adirect measure of ion relative abundance, but ionization effi-ciency can vary among different compound classes, so that ionrelative abundance does not necessarily reflect the relative abun-dance of the parent neutrals in the original sample.

All observed ions were singly charged, as evident from the unit m/z spacing between 12Cc and 13C12Cc-1 isotopic variants for each ele-mental composition. Mass values for each charged peak with FT-ICR mass spectral magnitude greater than 3r of baseline noise wereconverted from International Union of Pure and Applied Chemistry(IUPAC) measured mass to the Kendrick mass scale for whichCH2 = 14.0000 rather than 14.01565 Dalton (Da) according to Eq.(4) (Kendrick, 1963):

Kendrick mass ¼ IUPAC mass� ð14:00000=14:01565Þ ð4Þ

The advantage of Kendrick mass is that compounds of a homol-ogous series (i.e. having the same heteroatomic composition andnumber of rings plus double bonds, but differing number of –CH2

groups) will have similar KMD values (Kendrick, 1963). These val-ues were calculated from Eq. (5):

KMD ¼ ðnominal Kendrick mass� Kendrick massÞ ð5Þ

Nominal Kendrick mass was calculated by rounding the Kend-rick mass up to the nearest integer. Each peak was assigned achemical formula by use of a molecular formula calculator pro-gram based on mass value. The program was limited to compoundscomprised of up to 100 12C, 200 1H, 4 14N, 20 16O, 4 32S and 4 13C.For the low m/z peaks (<480 Da), only one possible chemical for-mula matched within 1 ppm of the measured mass. Identificationof the higher m/z compounds (>480 Da) often resulted in two ormore possible choices and results from KMD analysis were utilizedto determine the chemical formula (Hughey et al., 2001; Stensonet al., 2003).

Van Krevelen diagrams were applied to visualize similaritiesand differences for the complex ultrahigh resolution mass spectralcharacterization of DOM extracted from the water of the sixstreams (van Krevelen, 1950). The O:C atomic ratio versus theH:C atomic ratio was plotted for each assigned elemental composi-tion (Hughey et al., 2001). This type of analysis allows visualizationof the recovered compounds in relationship to important DOMcompound class, such as lignins, tannins, condensed hydrocarbons,cellulose and proteins (Kim et al., 2003a, 2004).

2.7. Statistical analysis

Fully nested ANOVA analysis with Tukey’s Honestly SignificantDifference (HSD) (p < 0.05) was performed on each chemical andphysical parameter to determine significant differences amongthe parameters by substrate type (limestone vs. sandstone) (Mini-tab 14.13). One-way ANOVA with Tukey’s HSD (p < 0.05) was usedto determine differences between DOM compound groups by sub-strate type (limestone vs. sandstone) (Minitab 14.13). Jaccard sim-ilarity coefficient analysis was used to determine similarities ofDOM compounds among the streams based on presence/absenceof raw m/z data (Statistica 6.0). Pearson correlation coefficientdetermined colinearity among the environmental parameters andcorrelated DOM compound group relative abundance (compoundgroup identification based strictly on plotting to the a region with-in the van Krevelen diagram) to stream water characteristics (SPSS14.0).

Constrained ordination techniques were utilized to identify pat-terns of variation in DOM quality among streams, correlations be-tween DOM quality and environmental descriptors, and theresponse variables contributing to the observed trends. Since cur-rently available multivariate statistical packages (Systat, SPSS,Canoco, etc.) do not support analysis of 2950–3700 response vari-ables, DOM quality data were grouped into compound ‘‘classes”(vs. compound groups when identified by use of van Krevelen dia-grams) based on the contribution of O, N and S to the molecularformula (i.e. O7, N2O2, or O8S1). Peak abundances for each classwere converted into weight percentage values by dividing by thetotal abundance for that sample; weight percentage values werenatural log transformed (ln + 1). Detrended correspondence analy-sis (DCA), an indirect gradient analysis based on segment length,was performed to determine the modality of the DOM quality dataand environmental predictor variables. The analysis resulted inshort (<1.0) segment lengths, indicating that the dataset was con-sidered linear and suitable for direct gradient analysis. Redundancyanalysis (RDA) and partial redundancy analysis (pRDA) were ap-plied (Canoco 4.5). In the RDA analysis, response variables weregrouped DOM quality data and predictor variables were measuredenvironmental parameters including acid neutralizing capacity,conductance, stream discharge, pH, slope, temperature, total sus-pended solids, watershed area, stream water concentrations ofAl3+, Ba2+, Ca2+, Cl�, dissolved inorganic carbon, dissolved oxygen,dissolved organic carbon, Fe2+,3+, K+, Mg2+, Mn2+,3+,4+, Na+, NHþ4 –N,NO3–N, Si2+,4+ and SO2�

4 , sediment chlorophyll a, total sedimentmicrobial biomass, sediment surface area and the percent canopy

Page 5: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188 1181

coverage. Forward selection of the predictor variables followed byMonte Carlo permutation tests were used to prevent artificial infla-tion of variation due to autocorrelation in the constrained ordina-tion model (Leps and Smilauer, 2003). The pRDA analysis wasperformed identically to that for the RDA except that stream loca-tion was used as a covariable. For the six-stream comparison, loca-tion was determined as the distance in kilometers from thewestern edge of the Bankhead National Forest to the stream mea-sured at 34.1225�N latitude (Fig. 1). For the among watershed com-parison, location was entered as 1 for all Bankhead streams and 2, 3and 4 for McDonald Branch, White Clay Creek and Rio Tempisquito,respectively.

2.8. Among watershed comparison

Previously published DOM quality data from three streams lo-cated in distant biomes were added to the DOM quality data pre-sented in this study and were compared by RDA and pRDA toplace the variability of DOM quality observed in our study withinthe variability observed at larger spatial scales. Data were fromMcDonald Branch located in the New Jersey Pinelands (Kim et al.,2003b), White Clay Creek in the Pennsylvania Piedmont and RioTempisquito located in a tropical rain forest biome in Costa Rica(Kim et al., 2006). A limited set of chemical characteristics of thestreams (conductivity, pH, Cl�, SO2�

4 , NO3–N, Ca2+, Na+, K+ andMg2+) was available from Findlay et al. (2008) The DOM from thesesamples was extracted by the same C18 extraction method andanalyzed at the same ESI FT-ICR MS facility. We applied the samedata reduction technique developed in this study to these profiles.

3. Results

3.1. Physical, chemical and biological parameters

Water chemistry of streams reflected the nature of the bedrockover which the streams flowed. Limestone streams had signifi-cantly higher concentrations of Ca2+, DIC, Mg2+, Na+ and SO2�

4 andsignificantly higher conductivity, ANC and pH values than thoseof the sandstone based streams (Table 1). Sandstone streams hadsignificantly higher concentrations of Al3+ and Fe2+,3+ comparedto limestone streams. We found no significant differences betweenthe remainder of the chemical parameters, biological characteris-tics or any of the physical parameters.

3.2. DOM characterization

A total of 5672 unique low molecular weight chemical formulae(250–1000 Da) were detected in stream water DOM through ultra-high resolution mass spectrometry. Of these, 960 (16.9%) werefound in all six. The percentage of chemical formulae found in 5,4, 3 and 2 streams were 14.9%, 15.0%, 13.8% and 19.7%, respec-tively. The remaining chemical formulae were distributed amongthe six streams (19.7%) but were unique to a particular stream.Presence/absence analysis of DOM quality revealed two groups ofsamples, the three Thompson Creek replicates and Hubbard, Flan-nagin and Borden Creeks samples, and the Brushy and Beech creekssamples (Table 2).

Van Krevelen diagrams indicated that all samples were rich incompounds found in the same regions as condensed hydrocarbonsand lignin/tannin derived compounds and lacking in lipids, pro-teins and cellulose (Fig. 2). Estimates of compound groups fromthe atomic O:C and atomic H:C ratios indicated that stream waterDOM from limestone streams had higher percentages of com-pounds found in the same region of the van Krevelen diagram ascondensed hydrocarbons than the sandstone streams, while sand-

stone stream water contained higher percentages of compoundsfound in the same region of the van Krevelen diagram as tannin de-rived compounds (Table 3). In addition, the average molecularweight of compounds found in the sandstone streams was higherthan those found in the limestone streams (Table 3). The relativeabundance of DOM compound groups correlated significantly withseveral water chemistry parameters, including condensed hydro-carbon-like compounds and the concentration Mg2+ and Ca2+

(r = 0.72 and r = 0.87, respectively) and tannin-like compoundsand the concentration of Al3+ and Fe2+,3+ (r = 0.73, r = 0.76,respectively).

3.3. DOM molecular class analysis

In addition to characterizing DOM quality with van Krevelendiagrams and similarity indices, we also characterized the DOMby ion classes described by heteroatom composition (subsequentlyreferred to as DOM compound classes while molecular formulaeidentified as similar by van Krevelen analysis are referred to asDOM compound groups; Appendix A). The most abundant classesof compounds found in the stream waters were O7, O8 and O9.These compounds ranged in abundance from 9.3% to 16.0% of thecompounds in each sample. In total, we found 35 compound clas-ses in both the limestone and sandstone streams and the relativeabundance of those classes served as the DOM quality descriptionused in subsequent multivariate analyses.

3.4. Correlation between DOM quality and environmental variables

RDA indicated that location was a significant environmentalvariable in describing the variation in DOM quality; therefore, weperformed a pRDA with location as a covariable. pRDA canonicalaxes 1 and 2 described 56.2% and 10.4% of the variation in DOMquality, respectively. The covariable, stream location, explainedan additional 15.3% of the variation in DOM quality (p = 0.008;F = 4.20). The pRDA triplot of samples, compound classes and envi-ronmental variables yielded two groups of samples along pRDAaxis one; sandstone streams had negative scores and limestonestreams had positive scores (Fig. 3). Thompson, the limestonestream with moderate levels of conductivity, pH, ANC, Ca2+ andNa+ concentrations was slightly positive, whereas the two streams(Borden and Flannagin creeks) with higher levels of conductivity,pH, ANC, Ca2+ and Na+ concentrations were more strongly positive.Two environmental factors, concentrations of Ca2+ (r = 0.7304) andAl3+ (r = �0.9546) correlated significantly with variation in DOMquality along pRDA axis 1. Ca2+ concentration correlated withDOM quality of the limestone streams and had greater abundancesin those streams. Al3+ concentration correlated with DOM qualityof the sandstone streams and was found in greater abundance inthose streams. Many of the measured parameters in this studywere autocorrelated. Forward selection of the environmental vari-ables followed by Monte Carlo permutations during pRDA selectedone member that had the best fit from among the correlated vari-ables. For example, Ca2+ concentration correlated with otherparameters found to be higher in limestone streams. These in-cluded ANC, pH, conductivity and the concentrations of DIC,Mg2+, SO2�

4 and Na+. Similarly, Al3+ concentration, which was high-er in the sandstone streams correlated with Fe2+,3+ and Mn2+,3+,4+

concentrations, chlorophyll a, total microbial biomass and TSS.pRDA triplot also identified compound classes that were signif-

icantly associated with variation in DOM quality among the sixstreams (Fig. 3); compound classes positively correlated withstreams showed higher relative abundance in those streams. TheseDOM compound classes were analyzed to determine the DOMcompounds groups by use of the algorithms of Kim et al. (2006)to determine which groups were present in each class. The classes

Page 6: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

Table 1Physical, chemical and biological characteristics of the Bankhead National Forest streams. Error estimates are represented by the standard error of the mean.

Parameter Borden Flannagin Thompson Beech Brushy Hubbard Significancea

Stream type Limestone Limestone Limestone Sandstone Sandstone Sandstone NALocationb (km�1) 14.50 13.50 6.00 20.00 23.00 2.50 NAAl3+ (mg L�1) 0.03 ± 0.00 0.04 ± 0.00 0.03 ± 0.00 0.06 ± 0.00 0.05 ± 0.00 0.05 ± 0.00 0.05ANC (meq L�1) 1.45 ± 0.01 1.78 ± 0.02 0.88 ± 0.02 0.10 ± 0.00 0.10 ± 0.02 0.07 ± 0.04 0.01Ba2+ (mg L�1) 0.02 ± 0.00 0.02 ± 0.00 0.02 ± 0.00 0.02 ± 0.00 0.02 ± 0.00 0.02 ± 0.00 NSCa2+ (mg L�1) 30.25 ± 0.03 34.78 ± 0.22 14.97 ± 0.00 1.69 ± 0.01 0.96 ± 0.01 1.21 ± 0.00 0.05Canopy (%) 40.04 ± 19.48 6.50 ± 4.52 1.29 ± 1.56 48.10 ± 31.39 93.34 ± 8.84 46.02 ± 31.50 NSChlorophyll a (lg g�1) 1.38 ± 0.30 2.31 ± 0.36 2.90 ± 0.29 7.42 ± 0.95 3.70 ± 1.95 3.76 ± 1.34 NSCl� (mg L�1) 1.11 ± 0.01 1.05 ± 0.01 1.10 ± 0.01 1.00 ± 0.01 0.85 ± 0.01 1.12 ± 0.01 NSConductance (lO) 157.17 ± 3.46 179.13 ± 0.12 49.00 ± 0.61 24.80 ± 0.10 17.10 ± 0.60 19.50 ± 0.52 0.05DIC (mg L�1) 0.37 0.43 0.22 0.03 0.02 0.20 0.01Discharge (m3 s�1) 0.59 0.19 0.51 0.32 0.29 0.42 NSDO (mg L�1) 9.68 ± 0.06 9.39 ± 0.18 9.63 ± 0.24 9.39 ± 0.11 9.69 ± 0.01 10.17 ± 0.01 NSDOC (mg L�1) 0.97 ± 0.43 0.92 ± 0.47 0.60 ± 0.01 1.31 ± 0.96 0.68 ± 0.19 0.89 ± 0.14 NSFe2+,3+ (mg L�1) 0.04 ± 0.00 0.04 ± 0.00 0.05 ± 0.00 0.16 ± 0.00 0.09 ± 0.00 0.09 ± 0.00 0.05K+ (mg L�1) 0.60 ± 0.01 0.59 ± 0.02 0.88 ± 0.00 0.62 ± 0.01 0.50 ± 0.01 0.74 ± 0.01 NSMg2+ (mg L�1) 1.63 ± 0.00 2.02 ± 0.01 1.49 ± 0.00 0.79 ± 0.00 1.04 ± 0.01 0.82 ± 0.01 0.01Mn2+,3+,4+ (mg L�1) 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00 0.03 ± 0.00 0.01 ± 0.00 0.01 ± 0.00 NSNa+ (mg L�1) 1.18 ± 0.01 1.17 ± 0.01 1.24 ± 0.00 1.10 ± 0.01 0.90 ± 0.00 0.93 ± 0.01 0.05NH4–N (lg L�1) 4.49 ± 0.19 0.80 ± 1.12 3.03 ± 1.22 4.32 ± 2.26 2.95 ± 1.75 3.93 ± 1.40 NSNO3–N (lg L�1) 10.70 ± 0.25 4.61 ± 0.49 44.31 ± 0.98 88.23 ± 47.30 2.67 ± 0.04 256.44 ± 23.60 NSpH 8.00 ± 0.00 8.07 ± 0.06 7.60 ± 0.00 7.17 ± 0.06 7.43 ± 0.21 7.13 ± 0.06 0.05Si2+,4+ (mg L�1) 2.92 ± 0.03 2.96 ± 0.01 2.80 ± 0.00 2.88 ± 0.02 2.99 ± 0.02 2.60 ± 0.02 NSSlope (%) 15.60 8.10 3.21 2.60 3.70 3.90 NA

SO2�4 (mg L�1) 4.26 ± 0.06 4.90 ± 0.07 4.35 ± 0.06 3.05 ± 0.02 3.67 ± 0.14 3.45 ± 0.31 0.05

Sediment surface area 0.70 ± 1.09 1.84 ± 0.67 2.30 ± 0.05 0.06 ± 0.01 0.74 ± 0.74 0.11 ± 0.02 NS(M3 g�1)Temp (C) 12.43 ± 0.06 13.30 ± 0.00 14.80 ± 0.30 12.24 ± 0.05 11.73 ± 0.06 12.50 ± 0.10 NSTotal Microbial Biomass 6.46 ± 0.42 9.85 ± 0.13 13.55 ± 1.33 20.03 ± 1.36 13.45 ± 5.58 25.31 ± 10.92 NS(nmol g�1)TSS (mg L�1) 3.15 ± 0.64 1.90 ± 0.00 0.75 ± 0.35 4.60 ± 2.55 1.70 ± 0.00 1.20 ± 0.28 NSWatershed area (km2) 26.06 19.94 10.62 12.66 13.75 9.63 NA

a Significant difference between substrate type (limestone vs. sandstone) detected by fully nested ANOVA; NA, Statistics not applied; NS, No significant difference.b As measured from the western edge of the forest; used as the covariable in pRDA analysis of Bankhead stream DOM quality.

Table 2Results of the Jaccard similarity coefficient defining similarities of DOM quality composition among streams in the Bankhead National Forest based on presence/absence data.Greater values indicate greater similarities.

Thompson 1 Thompson 2 Thompson 3 Flannagin Borden Beech Brushy Hubbard

Thompson 1 1.000 0.739 0.774 0.632 0.533 0.403 0.415 0.625Thompson 2 1.000 0.801 0.694 0.489 0.401 0.489 0.625Thompson 3 1.000 0.699 0.554 0.416 0.425 0.667Flannagin 1.000 0.502 0.422 0.434 0.742Borden 1.000 0.303 0.316 0.467Beech 1.000 0.617 0.445Brushy 1.000 0.460Hubbard 1.000

1182 J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188

representing the variation in the sandstone streams were com-prised predominately of lignin/tannins whereas the limestonestream classes were comprised of a mixture of condensed hydro-carbons and molecules not classified by the van Krevelen diagrams(Table 4).

3.5. Among watershed comparison

DOM quality data for stream water from Bankhead NationalForest, White Clay Creek, Rio Tempisquito and McDonald Branchwere compared by RDA with forward selection of environmentalvariables and the triplot showed DOM extracted from BankheadNational Forest streams differed in composition from that of theother streams (Fig. 4). RDA axis 1 described 31.5% of the variationbetween the samples and RDA axis 2 described 26.9% of the varia-tion (F = 4.87, p = 0.008). Samples from Bankhead National Forest

had negative scores whereas the samples from White Clay Creek,Rio Tempisquito and McDonald Branch had positive scores forRDA axis 1. Stream location was most strongly associated withthe variation (r = 0.8581) as was pH (r = -0.6727) and Na+

(r = 0.6987). The triplot also indicated that the DOM from Bank-head National Forest streams was enriched in compound classescontaining sulfur, whereas samples from the other streams wereenriched in nitrogen and/or oxygen compounds. Dissolution of or-ganic molecules from the shallow sulfur rich bituminous coal fieldfound throughout the Bankhead National Forest likely accounts forthese differences.

To further investigate the influence of bedrock type on streamwater DOM quality we performed a pRDA with location as a covar-iable on the data from the nine streams. pRDA axes 1 and 2 de-scribed 29.0% and 8.1% of the variation, respectively. Thecovariable, stream location explained 24.0% of the variation

Page 7: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

Brushy

Atomic O:C Ratio

Ato

mic

H:C

Rat

io

A

B CD

E

F

0

0.5

1

1.5

2

2.5

0 0.25 0.5 0.75 1

Flannagin

Atomic O:C Ratio

Ato

mic

H:C

Rat

io

A

B CD

E

F

0

0.5

1

1.5

2

2.5

0 0.25 0.5 0.75 1

Hubbard

Atomic O:C Ratio

Ato

mic

H:C

Rat

io

A

B CD

E

F

0

0.5

1

1.5

2

2.5

0 0.25 0.5 0.75 1

Thompson

Atomic O:C Ratio

Ato

mic

H:C

Rat

io

A

B CD

E

F

0

0.5

1

1.5

2

2.5

0 0.25 0.5 0.75 1

Beech

Atomic O:C Ratio

Ato

mic

H:C

Rat

io

A

B CD

E

F

0

0.5

1

1.5

2

2.5

0 0.25 0.5 0.75 1

Borden

Atomic O:C Ratio

Ato

mic

H:C

Rat

io

A

B CD

E

F

0

0.5

1

1.5

2

2.5

0 0.25 0.5 0.75 1

Fig. 2. Van Krevelen diagrams (O:C atomic ratio vs. H:C atomic ratio) constructed from ultrahigh resolution mass spectra of DOM extracted from stream water of threelimestone streams (Borden, Flannagin, and Thompson Creeks) and three sandstone streams (Beech, Brushy, and Hubbard Creeks) located in the Bankhead National Forest, AL.The range of O:C atomic ratios versus the H:C atomic ratios for several important DOM compound groups are indicated by: A = lipids, B = condensed hydrocarbons, C = lignin,D = tannins, E = proteins and F = cellulose (structural assignments per Kim et al., 2003a). A single, randomly chosen replicate was used to represent Thompson Creek.

Table 3Average molecular weight and percent contribution of compounds found in the same regions of the van Krevelen diagram as condensed hydrocarbon, lignin, tannin, protein, andcellulose to DOM extracted from stream water of Bankhead National Forest streams.

Borden Flannagin Thompson Brushy Beech Hubbard Significancea

Average size (Da) 468.75 505.79 499.43 520.20 522.23 524.50 0.05Condensed Hydrocarbon region (%) 22.54 20.61 20.81 19.03 15.63 11.35 0.01Lignin region (%) 5.8 15.09 12.42 9.72 27.67 16.73 NSTannin region (%) 13.67 15.55 12.83 27.54 29.62 17.19 0.01Protein region (%) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 NSCellulose region (%) <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 NS

a Significant differences between substrate type (limestone vs. sandstone) detected by one-way ANOVA.

J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188 1183

(F = 2.28, p = 0.036). pH (r = 0.8830) and Na+ (r = 0.7003) were sig-nificantly associated with the variation (Fig. 5). The triplot groupedstream water DOM quality samples by bedrock type and/or waterchemistry values.

After the removal of the variation associated with stream loca-tion, the DOM quality of the sandstone streams was similar to the

DOM quality of McDonald Branch. McDonald Branch flows througha cedar bog that overlays sediment characterized by CohanseySand, a Miocene aged coastal plain deposit (Bernard, 1963). TheDOM quality of the stream water in the limestone streams wasthe most similar to the DOM quality of White Clay Creek, whichis underlain by Cockeysville Marble, mafic gneiss and schist

Page 8: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

-1.0 1.5

-1.01.0

N2O12

O10S1

O11

O13

O3

O4S1

O5S1

O7S1

O9S1

Ca

Al

Location15.3%

pRDA Score 156.2%

pRD

A Score 210.4%

Fig. 3. Triplot of partial redundancy analysis (pRDA) of DOM quality of water fromBankhead National Forest streams with forward selection of predictor variables andstream location as a covariable. Solid arrows represent predictor variablessignificantly correlated to the variation in DOM quality, dotted arrows representDOM compound classes significantly associated with the variation betweensamples. The length of the arrow correlates with the degree of relation betweenthe response variable (DOM quality) and the associated predictor variable. Thearrows point in the direction of maximum change for the associated variable.Limestone streams (Borden, circle; Flannagin, square; Thompson, triangles) areindicated with filled symbols and sandstone streams (Beech, circle; Brushy, square;Hubbard Creek, triangle) are indicated with open symbols. Each symbol representsa single, independent sample. The percentage of total variation explained by eachaxis is listed below the axis label and the percentage explained by the covariable isgiven in the insert box.

Table 4Group composition of classes describing variation in the pRDA of samples extractedfrom stream water of Bankhead National Forest streams.

Compound class Streamsa Compound groupb

O3 Limestone Condensed hydrocarbon (100%)

O4S1 Limestone Condensed hydrocarbon (22.22%)Unclassified (77.78%)

O5S1 Limestone Condensed hydrocarbon (27.83%)Unclassified (72.17%)

O7S1 Limestone Condensed hydrocarbon (40.11%)Lignin/Tannin (9.34%)Unclassified (50.55%)

O9S1 Limestone Condensed hydrocarbon (13.37%)Lignin/Tannin (48.63%)Unclassified (48%)

N2O12 Limestone Condensed hydrocarbon (2.15%)Tannin/Lignin (55.60%)Unclassifieda (42.25%)

O11 Sandstone Condensed hydrocarbon (8.44%)Tannin/Lignin (81.57%)Unclassified (9.99%)

O13 Sandstone Condensed Hydrocarbon (3.91%)Lignin/Tannin (73.80%)Unclassified (22.29%)

O10S1 Sandstone Condensed hydrocarbon (79.50%)Tannin/Lignin (18.50%)Unclassified (2.00%)

a Stream type listed showed greater relative abundance of the compound class incollected DOM.

b Compound group names do not indicate molecular structure; rather theyindicate the members of a compound class that are found in the same region as thecompound group in the van Krevelen diagram. Unclassified indicates that vanKrevelen diagrams do not classify a percentage of the molecules in the compoundclass into known groups (i.e. these compounds plot outside of the boxes seen inFig. 1).

-1.0 1.0

-1.01.0 N2O5

N2O6

N2O7

O11S1

O12

O12S1

O13

O15O16

O9S1

pH

Na2+

Location

RDA Score 131.5%

RD

A Score 226.9%

Fig. 4. Triplot of redundancy analysis (RDA) of DOM quality data of water fromBankhead National Forest streams, White Clay Creek, Rio Tempisquito andMcDonald Branch with forward selection of predictor variables. Data presentationis as in Fig. 2, except that White Clay Creek, Rio Tempisquito and McDonald Branchsamples are represented by the cross-hatched circle, square and triangle,respectively.

1184 J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188

(Alcock, 2005). The DOM quality of Rio Tempisquito stream water,which flows over bedrock of volcanic origin (Newbold et al., 1995),was the most similar in DOM quality to that of White Clay Creekand the Bankhead National Forest limestone streams. We also ob-served that the water chemistry values from the three streamsintroduced to this study were similar according to geology.McDonald Branch had lower pH (pH 3.8–4.4) and conductivity val-ues (31–78 ls�1 cm) than White Clay Creek (pH 6.9–8.6, conduc-tivity 123–218 ls�1 cm) and Rio Tempisquito (pH 7.4–9.2 andconductivity 59–142 ls�1 cm). These values coincide with thewater chemistry values observed in the streams located in theBankhead National Forest, because the stream water of the sand-stone based streams have lower conductivity (sandstone 17–25 ls�1 cm vs. limestone 49–157 ls�1 cm) and pH (sandstone7.1–7.4 vs. limestone 7.6–8.0) values than the limestone basedstreams.

The triplot identified individual classes of compounds associ-ated with the variation of DOM quality between the streams. Theclasses representing the variation in the sandstone streams andMcDonald Branch were comprised predominately of lignin/tannintype compounds. The classes describing the variation associatedwith the limestone streams and White Clay Creek were predomi-nately comprised of condensed hydrocarbon classes (Table 5).

The triplicate samples taken in from Thompson Creek weremore similar to each other than any other samples (Table 2, Figs. 3–5) indicating that the extraction and characterization techniquesused in this study (e.g. C18 and ESI FT-ICR MS) are reproduciblewith low analytical error.

4. Discussion

Expressing DOM quality as the relative percent of identifiedcompound classes allowed for statistical comparison of DOM

Page 9: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

-1.0 1.0

-0.41.0

O1

O12

O13

O14

O3

O6

pH

Na

pRDA Score 129.0%

pRD

A Score 28.1%

Location24.0%

Fig. 5. Triplot of the partial redundancy analysis (pRDA) of DOM quality datacomparing samples taken from the Bankhead National Forest, White Clay Creek,Rio Tempisquito and McDonald Branch with forward selection of predictorvariables, and stream location as a covariable. Data presentation is as in Figs. 3and 4.

Table 5Group composition of classes describing variation in the pRDA of samples extractedfrom stream water of the six streams located in the Bankhead National Forest andWhite Clay Creek, McDonald Branch and Rio Tempisquito.

Compound class Streamsa Compound groupb

O1 High pH Condensed hydrocarbon (100%)

O3 High pH Condensed hydrocarbon (100%)

O6 High pH Condensed hydrocarbon (54.39%)Tannin/Lignin (0.39%)Unclassified (45.22%)

O12 Low pH Condensed hydrocarbon (4.8%)Tannin/Lignin (64.61%)Unclassified (30.59%)

O13 Low pH Condensed hydrocarbon (3.91%)Lignin/Tannin (73.8%)Unclassified (22.29%)

O14 Low pH Lignin/Tannin (42.96%)Unclassified (57.04%)

a Stream type listed showed greater relative abundance of the compound class incollected DOM.

b Compound group names do not indicate molecular structure; rather theyindicate the members of a compound class that are found in the same region as thecompound group in the van Krevelen diagram. Unclassified indicates that vanKrevelen diagrams do not classify a percentage of the molecules in the compoundclass into known groups (i.e. these compounds plot outside of the boxes seen inFig. 1).

J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188 1185

quality. Although van Krevelen diagrams are a useful tool forcomparing major differences in molecular composition betweena small number of samples, it is difficult to quantify the degreeof difference among samples or determine compounds responsi-ble for the variation. In addition, the van Krevelen based com-pound groups account for less than 60% of the compoundsidentified by ESI FT-ICR MS. Our approach has made a significantadvance in the study of DOM quality by utilizing the algorithmdeveloped in the MIDAS software to produce a 35 membermultivariate description of DOM quality. Using this approachwe were able to identify differences in DOM quality amongstreams, identify specific DOM compound classes associated withthese differences and the environmental parameters that corre-lated with the differences. Those tested included various mea-sures of stream water chemistry, stream discharge, slope,temperature, total suspended solids, watershed area, sedimentchlorophyll a, total sediment microbial biomass, sediment surfacearea and the percent canopy coverage. The comparison betweenJaccard similarity coefficient results (ungrouped presence/ab-sence data) and RDA (grouped relative percentage data) indicatedthat our method of grouping the compounds did not change ourconclusions as both analyses grouped the samples in a similarfashion.

To the best of our knowledge, this study is the first to combineESI FT-ICR MS and multivariate statistical analysis to explore theinfluence of watershed geology on DOM quality. Our results indi-cated that the bedrock over which streams flow influences thequality of DOM in stream water. These differences were correlatedto pH and cation concentrations in the stream water. Streams flow-ing over sandstone bedrock consistently showed lower pH and cat-ion concentration and higher relative abundance of DOMcompound classes comprised of molecules that plotted with con-densed hydrocarbons in van Krevelen diagrams (either partiallyor exclusively). Streams flowing over limestone bedrock showedhigher pH and cation concentrations and higher relative abun-dance of DOM compound classes comprised of molecules that plot-ted with lignin/tannin and condensed hydrocarbons in vanKrevelen diagrams and molecules that were not classified intocompound groups by the traditional van Krevelen analysis(Table 5).

Cations form associations with the DOM compounds found inthe stream water and can cause flocculation. Many studies haveshown that Al3+ and Fe2+,3+ form strong ligand exchange bondswith aromatic DOM moieties (typical of condensed hydrocarbons)and have a higher binding affinity for aromatic DOM compoundsthan alkaline earth metals (Kawahigashi et al., 2006; Mikuttaet al., 2007). Interestingly, in our study, stream water that con-tained higher concentrations of Al3+ and Fe2+,3+ (i.e. sandstonestreams) also had higher relative abundance off molecules thatplotted with condensed hydrocarbons in van Krevelen diagrams.Ca2+ concentration, which is 100–500 fold greater in limestonestreams than Al3+ and Fe2+,3+ concentrations in the sandstonestreams, likely caused this deviation from the expected pattern.Cation concentration (particularly Ca2+) influences DOM bindingto mineral particle surfaces via a number of proposed mechanismsincluding Ca2+ serving as a bridge between negatively charged or-ganic solutes and negatively charged mineral surfaces, and Ca2+

complexation with negative functional groups of individual DOMmolecules leading to either decreased charge density or reducednumber of binding groups per molecule (Day et al., 1994; Guggen-berger and Kaiser, 2003). These bridges are important at alkalinepH and high Ca2+ concentrations as calcium ions have been shownto cause flocculation and subsequent precipitation of DOM byincreasing adsorption affinity of DOM to mineral surfaces (PingWeng et al., 2005). These Ca2+-DOM interactions, although not asstrong as those occurring between DOM and Al3+ and Fe2+,3+, com-bined with the approximate order of magnitude increase in cationconcentration in limestone streams, contribute to the greater rela-tive abundance of condensed hydrocarbons in the sandstonestreams.

Noticeably absent in our DOM quality characterizations arecompounds in excess of 1000 Da. Although it is possible thatbiases in either the C18 extraction procedure or the ESI MS couldaccount for the absence of these molecules (see Kim et al., 2006for a full discussion), the sorption properties of the terrigenoussoil and mineral horizons likely prevent the compounds fromreaching stream water. When DOM flows through a porous sub-strate, higher molecular weight DOM shows greater reactivity

Page 10: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

1186 J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188

and binds to the substrate, leaving lower molecular weight DOMmolecules to pass through (Meier et al., 2004; Neff et al., 2006).Another study found soils with higher clay content and higherspecific surface area had greater absorption potential for DOMmolecules than coarse sandy soils (Nelson et al., 1990). Thismay also explain the significant correlation between watershedarea and DOM quality as our sampling protocol (sampling14 days after last rain event) insured that the vast majority ofsampled stream water interacted with the forest soil and/or min-eral horizons below prior to entering the stream.

Stream location was the most significant factor associatedwith the variation in stream water DOM quality. We utilizedlocation as a synthetic variable to account for factors not mea-sured in this study that can influence DOM quality. Factorsshown to influence stream water DOM quality include: (1)sources of the organic matter (such as terrigenous plant commu-nities, detrital terrigenous carbon, fossilized carbon, autochtho-nous production), (2) processes altering allochthonous DOMquality during transport to the streams (i.e., precipitation, relief,soil and mineral horizon composition, litter fall and microbialdegradation in soils) and (3) in-stream processing such as photo-oxidation and microbial degradation (Aitkenhead-Peterson et al.,2003). Many of these factors vary seasonally (Maurice and Leff,2002). Our sampling protocol, site selection and experimental de-sign eliminated seasonal variation and differences in autochtho-nous production, precipitation and photooxidation. While thedifferences in riparian community composition (the presence ofT. canadensis in the riparian zones of the sandstone streams)could indirectly contribute to the observed changes in DOM qual-ity associated with bedrock geology, they do not explain theimportance of location in the statistical analysis of factors affect-ing DOM quality as stream type (sandstone vs. limestone)showed no correlation to location (Table 1). Similarly, we elimi-nate differences in detrital terrigenous carbon, soil and mineralhorizon composition and microbial degradation in soils as proxi-mal causes of the location based differences as the watershed ofeach stream drains the same forest, with soil derived from thesame bedrock (Pennsylvanian age Pottsville sandstone). Statisticalanalysis indicated that location did not correlate with variation instreambed microbial community structure (Mosher, 2008) sug-gesting that differences in in-stream microbial degradation ofDOM was not the source of the variation in DOM quality dueto location. What remains as the most likely candidate as thesource of the correlation between DOM quality and location isthe contribution of fossilized carbon to stream water DOM. Thehigh relative abundance of sulfur containing molecules in Bank-head stream water DOM (in excess of 10% for some streams)and the correlation (y = �0.15x + 8.95, R2 = 0.30) of the relativeabundance of sulfur containing molecules to stream location sup-port this conclusion. The coal bed located throughout the studyarea is intermittent and varies in thickness. Drill hole data forcoal content is available from two sites (Beech and Brushy creeks)and indicated a 0.5 m thick layer of bituminous coal in the Park-wood sandstone layer at each stream (Ward and Chase, 1981).The remaining four streams are located within the protected Sip-sey Wilderness Area and the quantities of coal at these sites havenot been determined but likely vary among sites. This unquanti-fiable variation required application of the partial redundancyanalysis to remove the effects of fossil carbon on stream waterDOM; its successful application demonstrates the utility of thisstatistical approach.

Interestingly, a previous report found that DOM quality instream waters was influenced through decomposition by soil mi-crobes whereas DOM concentration was influenced by soil sorp-tion properties as a result of watershed geology (Nelson et al.,

1993). In this current study, soil microbial communities and ter-rigenous physical characteristics, with the exception of wa-tershed area and slope, were not examined. Another possibilitythat could be attributed to the effects of location is subsurfacehydrology. Slight variation in landscape features, hillside posi-tioning, orientation, bedrock fractures or any combination there-in has been shown to affect hydrological flow (Lin and Zhou,2008).

Several factors correlated with variation in both the within andamong watershed comparisons. Again, most notable was theimportance of location. The importance of sulfur enriched com-pounds, shown by the RDA triplot (Fig. 4), in describing the varia-tion in DOM quality among watersheds indicated that variation infossilized carbon input to stream water DOM was responsible, inpart, for the among location differences. DOM quality also corre-lated with pH (and by autocorrelation with conductivity, Ca2+,Na+ and Mg2+ concentrations) in both analyses and once the effectsof stream location were removed the samples grouped according togeology/water chemistry parameters in both pRDAs (Figs. 3 and 5),further corroborating the importance of water chemistry parame-ters as determinants of DOM quality. Moreover, after the removalof location effects, several of the same compound classes describedthe variation in DOM quality in both the within and among wa-tershed comparisons, suggesting that these molecules were moststrongly affected by the direct and indirect effects of bedrockgeology.

5. Conclusions

Although our study design is correlative, the similarities be-tween the small and large spatial scale comparisons indicate thatstream water inorganic chemistry does, in part, influence DOMquality. The design did not allow differentiation of direct effects,through water chemistry, from indirect effects such as plant com-munity composition, soil and mineral horizon interactions or bio-logical processes in the forest soils. This study also providedstrong evidence that fossilized carbon in bedrock, when present,can also influence stream water DOM quality. Application andexpansion of the multivariate statistical analyses explored in thisstudy should allow future studies to further refine theserelationships.

Acknowledgements

The authors acknowledge Lou Kaplan, Pat Hatcher and Sungh-wan Kim for insightful discussions on data interpretation and gen-erously providing dissolved organic matter quality data and EricRoden for use of B.E.T. analyzer at the University of Wisconsin-Madison. Rusty Ward and Richard Carroll of the Alabama Geologi-cal Survey and Tom Counts of the National Forest Service providedtechnical advice. Lori Tolley-Jorden, Mike Chadwick and TimothyWynn helped collect samples. The Aquatic Chemistry Laboratory,Center for Freshwater Studies, The University of Alabama (Dr.Amelia K. Ward, Director, Anne Bell and Kevin Pritchard) per-formed the analyses of inorganic nutrient and dissolved organiccarbon concentrations.

This work was supported by the J. Nicholene Bishop endow-ment (RHF) and an Aquatic Biology Enhancement Fellowship(JJM; Department of Biological Sciences, The University of Ala-bama); the National Science Foundation (DMR 06-54118, DEB05-16235); the State of Florida; and made possible through theUnited States Department of Agriculture Forest Service SpecialUse Permit BAN201601.

Page 11: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188 1187

Appendix A

Relative abundance of compound classes in collected DOM of the six Bankhead National Forest streams.

Compound

Thompson 1 Thompson 2 Thompson 3 Flannagin Borden Beech Brushy Hubbard

O1

0.09 0.09 0.04 0.10 0.03 0.09 0.46 0.05 O2 0.48 0.14 0.53 0.34 0.29 0.28 1.45 0.12 O3 1.87 1.28 1.72 1.55 3.17 0.78 1.54 0.51 O4 6.80 3.50 4.48 3.96 7.11 1.47 2.68 2.88 O5 9.84 5.86 7.79 6.88 11.01 3.36 4.19 5.52 O6 12.02 9.85 10.76 10.71 13.60 6.08 6.47 9.38 O7 13.44 13.45 13.25 14.49 14.11 9.28 9.14 12.63 O8 11.74 15.22 14.69 15.99 12.04 13.14 12.79 14.13 O9 10.46 15.40 12.37 15.15 11.22 15.81 14.73 14.98 O10 8.11 11.19 10.04 12.60 8.42 15.79 15.07 12.36 O11 5.96 7.79 6.22 1.96 5.00 12.71 12.41 8.08 O12 4.34 5.23 4.25 5.83 2.38 8.37 8.24 6.07 O13 2.75 3.15 2.35 3.63 0.65 4.77 3.51 3.64 O14 0.79 1.70 1.08 2.11 0.15 1.92 1.52 1.60 O4S1 0.02 0.00 0.04 0.11 0.11 0.00 0.00 0.00 O5S1 0.27 0.12 0.29 0.12 0.38 0.00 0.00 0.00 O6S1 0.07 0.20 0.40 0.15 0.30 0.02 0.06 0.31 O7S1 0.16 0.27 0.49 0.11 0.80 0.11 0.30 0.17 O8S1 1.13 0.48 1.06 0.11 3.14 0.58 0.28 1.42 O9S1 2.22 0.71 2.07 0.43 2.29 1.25 1.44 1.03 O10S1 2.41 1.54 1.91 0.00 0.00 1.82 1.19 2.02 O11S1 2.08 1.22 1.96 2.05 1.45 1.00 0.75 1.80 O12S1 1.53 0.99 1.10 0.82 0.70 0.58 0.32 0.79 O13S1 0.70 0.33 0.44 0.41 0.15 0.23 1.20 0.21 O14S1 0.20 0.11 1.11 0.05 0.14 0.02 0.05 0.06 N2O5 0.02 0.01 0.02 0.00 0.01 0.02 0.01 0.00 N2O6 0.07 0.01 0.00 0.00 0.01 0.00 0.00 0.00 N2O7 0.05 0.03 0.00 0.00 0.01 0.00 0.00 0.00 N2O8 0.01 0.11 0.02 0.00 0.05 0.02 0.01 0.16 N2O9 0.02 0.01 0.02 0.01 0.11 0.00 0.00 0.00 N2O10 0.02 0.01 0.05 0.03 0.15 0.05 0.01 0.00 N2O11 0.02 0.02 0.08 0.03 0.19 0.00 0.00 0.00 N2O12 0.09 0.09 0.07 0.08 0.29 0.00 0.00 0.00 N2O13 0.30 0.03 0.08 0.09 0.11 0.04 0.01 0.00 N2O14 0.00 0.00 0.00 0.00 0.00 0.06 0.01 0.00

Associate Editor—Daniel Hunkeler

References

Aitkenhead-Peterson, J.A., McDowell, W.H., Neff, J.C., 2003. Sources, production, andregulation of allochthonous dissolved organic matter inputs to surface waters.In: Findlay, S.E.G., Sinsabaugh, R.L. (Eds.), Aquatic Ecosystems Interactivity ofDissolved Organic Matter. Academic Press, San Diego, pp. 25–70.

Alcock, J., 2005. Bedrock Geology of the West Grove Quadrangle, Pennsylvania-Delaware Piedmont. Geological Society of America Digital Map and Chart 002.doi: 10.1130/2005.DMCH002.

Bernard, J.M., 1963. Forest floor moisture capacity of the New Jersey Pine Barrens.Ecology 44, 574–576.

Brunauer, S., Emmet, P.H., Teller, E., 1938. Adsorption of gases in multimolecularlayers. Journal of the American Chemical Society 60, 309–319.

Day, G.M., Hart, B.T., McKelvie, I.D., Beckett, R., 1994. Adsorption of natural organicmatter onto goethite. Colloids and Surfaces A: Physicochemical and EngineeringAspects 89, 1–13.

Diehl, S.F., Goldhaber, M.B., Hatch, J.R., 2004. Modes of occurrence of mercury andother trace elements in coals from the warrior field, Black Warrior Basin,Northwestern Alabama. International Journal of Coal Geology 59, 193–208.

Fanson, M.A., 1985. Standard Methods for the Examination of Water andWastewater, nineteenth ed. American Public Health Association.

Findlay, R.H., King, G.M., Watling, L., 1989. Efficacy of phospholipid analysis indetermining microbial biomass in sediments. Applied and EnvironmentalMicrobiology 55, 2888–2893.

Findlay, R.H., Yeates, C., Hullar, M.A.J., Stahl, D.A., Kaplan, L.A., 2008. Biome-levelbiogeography of streambed microbiota. Applied and EnvironmentalMicrobiology 74, 3014–3021.

Frost, P.C., Larson, J.H., Johnston, C.A., Young, K.C., Maurice, P.A., Lamberti, G.A.,Bridgham, S.D., 2006. Landscape predictors of stream dissolved organic matterconcentration and physiochemistry in a Lake Superior river watershed. AquaticSciences 68, 40–51.

Guggenberger, G., Kaiser, K., 2003. Dissolved organic matter in soil: challenging theparadigm of sorptive preservation. Geoderma 113, 293–310.

Hughey, C.A., Hendrickson, C.L., Rodgers, R.P., Marshall, A.G., Qian, K., 2001.Kendrick mass defect spectrum: a compact visual analysis for ultrahigh-resolution broadband mass spectra. Analytical Chemistry 73, 4676–4681.

Kaplan, L.A., Newbold, J.D., 1993. Biogeochemistry of dissolved organic carbonentering streams. In: Ford, T.E. (Ed.), Aquatic Microbiology: An EcologicalApproach. Blackwell, New York, pp. 139–165.

Kawahigashi, M., Kaiser, K., Rodionov, A., Guggenberger, G., 2006. Sorption ofdissolved organic matter by mineral soils of the Siberian forest tundra. GlobalChange Biology 12, 1868–1877.

Kendrick, E., 1963. A mass scale based on CH2 = 14.00000 for high-resolution massspectrometry of organic compounds. Analytical Chemistry 35, 2146–2154.

Kim, S., Kramer, R.W., Hatcher, P.G., 2003a. Graphical method for analysis of ultra-high resolution broadband mass spectra of natural organic matter, the vanKrevelen diagram. Analytical Chemistry 75, 5336–5344.

Kim, S., Simpson, A.J., Kujawinski, E.B., Freitas, M.A., Hatcher, P.G., 2003b. Highresolution electrospray ionization mass spectrometry and 2D solution NMR forthe analysis of DOM extracted by C18 solid phase disk. Organic Geochemistry34, 1325–1335.

Kim, S., Kaplan, L.A., Benner, R., Hatcher, P.G., 2004. Hydrogen-deficient moleculesin natural riverine water samples-evidence for the existence of black carbon inDOM. Marine Chemistry 92, 225–234.

Page 12: Influence of bedrock geology on dissolved organic matter ...crcooper01.people.ysu.edu/Mosher2010_OrganicGeochem_DOMQual… · Influence of bedrock geology on dissolved organic matter

1188 J.J. Mosher et al. / Organic Geochemistry 41 (2010) 1177–1188

Kim, S., Kaplan, L.A., Hatcher, P.G., 2006. Biodegradable dissolved organic matter ina temperate and a tropical stream determined from ultra-high resolution massspectrometry. Limnology and Oceanography 5, 1054–1063.

Koch, B.P., Witt, M., Engbrodt, R., Dittmar, T., Kattner, G., 2005. Molecular formulaeof marine and terrigenous dissolved organic matter detected by electrosprayionization Fourier transform ion cyclotron resonance mass spectrometry.Geochimica et Cosmochimica Acta 69, 3299–3308.

Kujawinski, E.B., Freitas, M.A., Zang, X., Hatcher, P.G., Green-Church, K.G., Jones, R.B.,2002. The application of electrospray ionization mass spectrometry (ESI MS) tothe structural characterization of natural organic matter. Organic Geochemistry33, 171–180.

Ledford Jr., E.B., Rempel, D.L., Gross, M.L., 1984. Space charge effects in Fouriertransform mass spectrometry. Analytical Chemistry 56, 2744–2748.

Leps, J., Smilauer, P., 2003. Multivariate Analysis of Ecological Data Using CANOCO.Cambridge University Press, Cambridge.

Lin, H., Zhou, X., 2008. Evidence of subsurface preferential flow using soil hydrologicmonitoring in the Shale Hills catchment. European Journal of Soil Science 59,34–49.

Magalhaes, M.F., Batalha, D.C., Collares-Pereira, M.J., 2002. Gradients in stream fishassemblages across a Mediterranean landscape: contributions of environmentalfactors and spatial structure. Freshwater Biology 47, 1015–1031.

Mancini, E.A., 1988. Geologic Map of Alabama. Geological Survey of Alabama,Special Map 220.

Marshall, A.G., Verdun, F.R., 1990. Fourier Transforms in NMR, Optical, and MassSpectrometry: A User’s Handbook. Elsevier, Amsterdam.

Marshall, A.G., Hendrickson, C.L., Jackson, G.S., 1998. Fourier transform ioncyclotron resonance mass spectrometry: a primer. Mass SpectrometryReviews 17, 1–35.

Mast, M.A., Turk, J.T., 1999. Environmental characteristics and water quality ofHydrologic Benchmark Network stations in the Eastern United States, 1963–1995. United States Geological Survey Circular 1173-A.

Maurice, P.A., Leff, L.G., 2002. Hydrogeochemical controls on the organic matter andbacterial ecology of a small freshwater wetland in the New Jersey Pine Barrens.Water Research 36, 2561–2570.

Maurice, P.A., Cabaniss, S.E., Drummond, J., Ito, E., 2002. Hydrogeochemical controlson the chemical characteristics of natural organic matter at a small freshwaterwetland. Chemical Geology 187, 59–77.

Meier, M., Chin, Y., Maurice, P., 2004. Variations in the composition and adsorptionbehavior of dissolved organic matter at a small, forested watershed.Biogeochemistry 67, 39–56.

Mikutta, R., Mikutta, C., Kalbitz, K., Scheel, T., Kaiser, K., Jahn, R., 2007.Biodegradation of forest floor organic matter bound to minerals viadifferent binding mechanisms. Geochimica et Cosmochimica Acta 71, 2569–2590.

Mosher, J.J., 2008. Geological influences on microbial community structure anddissolved organic matter quality in forested streams. Ph.D. Dissertation,University of Alabama, Tuscaloosa.

Neff, J.C., Reynolds, R., Sanford, R.L., Fernandez, D., Lamothe, P., 2006. Controls ofbedrock geochemistry on soil and plant nutrients in southeastern Utah.Ecosystems 9, 879–893.

Nelson, P.N., Cotsaris, E., Oades, J.M., Bursill, D.B., 1990. Influence of soil clay contenton dissolved organic matter in stream waters. Australian Journal of Marine andFreshwater Research 41, 761–764.

Nelson, P.N., Baldock, J.A., Oades, J.M., 1993. Concentration and composition ofdissolved organic carbon in streams in relation to catchment soil properties.Biogeochemistry 19, 27–50.

Newbold, D.N., Sweeney, B.W., Jackson, J.K., Kaplan, L.A., 1995. Concentrations andexport of solutes from six mountain streams in northwestern Costa Rica. Journalof the North American Benthological Society 14, 21–37.

Ping Weng, L., Koopal, L.K., Hiemstra, T., Meeussen, J.C.L., Van Riemsduk, W.H., 2005.Interactions of calcium and fulvic acid at the goethite–water interface.Geochimica et Cosmochimica Acta 69, 325–339.

Qian, K., Robbins, W.K., Hughey, C.A., Cooper, H.J., Rodgers, R.P., Marshall, A.G., 2001.Resolution and identification of elemental compositions for more than 3000crude acids in heavy petroleum by negative-ion microelectrospray high fieldFourier transform ion cyclotron resonance mass spectrometry. Energy and Fuels15, 1505–1511.

Schumacher, M., Christl, I., Vogt, R.D., Barmettler, K., Jacobsen, C., Kretzschmar, R.,2006. Chemical composition of aquatic dissolved organic matter in five borealforest catchments sampled in spring and fall seasons. Biogeochemistry 80, 263–275.

Senko, M.W., Hendrickson, C.L., Pasa-Tolic, L., Marto, J.A., White, F.M., Guan, S.,Marshall, A.G., 1996a. Electrospray ionization FT-ICR mass spectrometry at 9.4T. Rapid Communications in Mass Spectrometry 10, 1824–1828.

Senko, M.W., Canterbury, J.D., Guan, S., Marshall, A.G., 1996b. A high-performancemodular data system for FT-ICR mass spectrometry. Rapid Communications inMass Spectrometry 10, 1839–1844.

Senko, M.W., Hendrickson, C.L., Emmett, M.R., Shi, S.D.-H., Marshall, A.G., 1997.External accumulation of ions for enhanced electrospray ionization Fouriertransform ion cyclotron resonance mass spectrometry. Journal of the AmericanSociety for Mass Spectrometry 8, 970–976.

Shi, S.D.-H., Drader, J.J., Freitas, M.A., Hendrickson, C.L., Marshall, A.G., 2000.Comparison and interconversion of the two most common frequency-to-masscalibration functions for Fourier transform ion cyclotron resonance massspectrometry. International Journal of Mass Spectrometry 195 (196), 591–598.

Smoot, J.C., Langworthy, D.E., Levy, M., Findlay, R.H., 1998. Periphyton growth onsubmerged artificial substrate as a predictor of phytoplankton response tonutrient enrichment. Journal of Microbiological Methods 32, 11–19.

Stenson, A.C., Marshall, A.G., Cooper, W.T., 2003. Exact masses and chemicalformulas of individual Suwannee River fulvic acids from ultrahigh resolutionelectrospray ionization Fourier transform ion cyclotron resonance mass spectra.Analytical Chemistry 75, 1275–1284.

Szabo, M.W., Copeland Jr., C.W., 1988. Special Map 220. Geologic map of Alabama,northwest sheet. Geological Survey of Alabama.

van Krevelen, D.W., 1950. Graphical-statistical method for the study of structureand reaction processes of coal. Fuel 29, 269–284.

Ward II, W.E., Chase, D.D., 1981. Map of Selected Coal Beds in Parts of the WarriorCoal Basin and the Plateau Coal Region. Haleyville Quadrangle, Alabama. Map181 C. Geological Survey of Alabama.

Welch, K.A., Lyons, W.B., Graham, E., Neumann, K., Thomas, J.M., Mikesell, D., 1996.Determination of major element chemistry in terrigenous waters fromAntarctica by ion chromatography. Journal of Chromatography A 739, 257–263.

Wetzel, R.G., Likens, G.E., 2000. Limnological Analysis, 3rd ed. Springer, New York.Wilcox, B.E., Hendrickson, C.L., Marshall, A.G., 2002. Improved ion extraction from a

linear octopole ion trap: SIMION analysis and experimental demonstration.Journal of the American Society for Mass Spectrometry 13, 1304–1312.