lable at ScienceDirect

Soil Biology & Biochemistry 69 (2014) 398e405

Contents lists avai

Soil Biology & Biochemistry

journal homepage: www.elsevier .com/locate/soi lbio

Soil organic matter stability in organo-mineral complexes asa function of increasing C loading

Wenting Feng a, Alain F. Plante a,*, Anthony K. Aufdenkampe b, Johan Six c

aDepartment of Earth & Environmental Science, University of Pennsylvania, Hayden Hall, 240 South 33rd Street, Philadelphia, PA 19104-6316, USAb Stroud Water Research Center, Avondale, PA 19311, USAcDepartment of Environmental Systems Science, Swiss Federal Institute of Technology, 8092 Zurich, Switzerland

a r t i c l e i n f o

Article history:Received 25 June 2013Received in revised form25 November 2013Accepted 27 November 2013Available online 12 December 2013

Keywords:Carbon loadingDOM sorptionIncubationOrgano-mineral complexesStabilityThermal analysis

* Corresponding author. Fax: þ1 215 898 0964.E-mail address: aplante@sas.upenn.edu (A.F. Plant

0038-0717/$ e see front matter � 2013 Elsevier Ltd.http://dx.doi.org/10.1016/j.soilbio.2013.11.024

a b s t r a c t

Soil carbon (C) saturation behavior predicts that soil C storage efficiency observed under field conditionsdecreases as a soil approaches C saturation. This may be due to a decline in soil organic matter (SOM)stability as the result of changes in the type, strength or turnover time of organo-mineral interactionswith increasing organic C input. The goal of this study was to test whether the stability of organic matterbound to soil minerals decreases as organo-mineral complexes approach C saturation with increasing Cloading. A series of batch sorption experiments with natural dissolved organic matter (DOM) and soilmineral components was conducted to obtain organo-mineral complexes with a range of organic Cloadings. The relative stability of C in these organo-mineral complexes was subsequently assessed usingevolved CO2 gas analysis during thermal analyses and laboratory incubations. Results indicated thatdifferences in dissolved organic C before and after sorption overestimated the amount of sorbed C whencompared to differences in solid-phase C concentrations. Values of C/N, d13C, and d15N of the organo-mineral complexes were significantly smaller or more negative than initial soil samples or the stockDOM solution, consistent with the concept of molecular fractionation by sorption to minerals and sug-gesting that the composition of the organic matter in the organo-mineral complexes may have changedas the amount of sorbed organic matter increased. Observations that organic C loadings at maximumsorption did not substantially exceed 1 mg C m�2 and linear rather than asymptotic relations betweensorbed C and initial DOM concentrations suggest that organo-mineral complexes may not have reached Csaturation under the conditions in the batch sorption experiments of this study. The temperature atwhich half of the CO2 evolved during thermal analysis (i.e., CO2-T50) increased with increasing C loading,suggesting the sorbed C required greater energy input for combustion. Results of the laboratory in-cubations to determine relative biological stability of sorbed C were not consistent with the initial hy-pothesis. The size of potentially mineralizable C pool relative to the total sorbed C decreased withincreasing C loading, even though this pool was being degraded more rapidly. Overall, the results did notsupport the hypothesis that SOM stability decreases with increasing C loading. In spite of generating awide range of C loadings substantially greater than previous studies, the conditions of the DOM sorptionexperiments conducted in this study appeared unable to generate organo-mineral complexes exhibitingC saturation behavior. We speculate that measurable decreases in SOM stability may occur only once thethreshold of C saturation is reached.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Most current models of soil organic matter (SOM) dynamicsassume that soil organic carbon (C) stocks increase linearly withincreasing organic C inputs without an upper limit (Paustian et al.,1997). In a few cases, however, soil C stocks have showed no

e).

All rights reserved.

significant increase in response to increasing long-term organic Cinputs, reaching a maximal soil C level (Campbell et al., 1991; Guldeet al., 2008; Huggins et al., 1998; Soon,1998). This phenomenon hasbeen proposed as the soil C saturation concept (Six et al., 2002;Stewart et al., 2007; West and Six, 2007), and the maximal soil Cstock has been defined as the soil C saturation level. Subsequentstudies demonstrated that chemically protected organic matterbound to soil minerals as organo-mineral complexes are morelikely to reach C saturation than other physical or chemical

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405 399

fractions or the bulk soil (Chung et al., 2008; Gulde et al., 2008;Stewart et al., 2008b). The C saturation of mineral-bound organicmatter may be attributable to the finite amount of mineral specificsurface area (SSA) onto which organic matter may be stabilized(Hassink, 1996). The amount of organic C stabilized on soil mineralscan be expressed as soil C loading, which is the mass of mineral-bound C per unit of mineral surface area (i.e., mg C m�2). OtherSOM fractions as less likely to saturate because of the lack of limitsto the amount of particulate organic matter (POM) that can bephysically protection in intra-aggregate pores.

Organic C inputs are stabilized onto mineral surfaces throughvarious organo-mineral bonding reactions such as ligand exchange,cation bridging, H-bonding, and van der Waal forces, depending onthe composition of the organic inputs, soil mineralogy, and envi-ronmental factors (Arnarson and Keil, 2000; Gu et al., 1994;Stevenson, 1994). Organic C stabilization on soil minerals ac-counts for a largemajority of total soil organic C (Christensen, 1998;Kahle et al., 2002), and mineral-bound C has longer turnover timesthan other fractions such as POM or physically protected organic Cin aggregates (Balesdent et al., 1987, 1988; Trumbore, 2000).Changes in the amount or stability of organic C inputs stabilized onsoil minerals will therefore greatly affect bulk soil C storage in thelong term.

According to the soil C saturation concept, soil C storage effi-ciency decreases as a soil approaches C saturation (Stewart et al.,2007, 2008a). A decrease in soil C storage efficiency may be dueto a decline in SOM stability as soils approach C saturation, whichcould be the result of changes in the type, strength or turnover timeof organo-mineral interactions with increasing organic C inputs(Kleber et al., 2007; Sollins et al., 2009). That is, as C concentrationsand C loadings increase, weaker organiceorganic interactionsbecome relatively more abundant than stronger organo-mineralinteractions as suggested by the zonal model for SOM stabiliza-tion (Kleber et al., 2007). We therefore hypothesized that the sta-bility of organic C bound to soil minerals decreases with increasingsoil C loadings as the organo-mineral bonding strength changesfrom strong toweak. It is difficult to sample natural organo-mineralcomplexes from field soil samples with well constrained miner-alogy and with a wide range of organic C loadings reaching highvalues (much greater 1 mg C m�2). Hence, we conducted a series oflaboratory sorption experiments with natural dissolved organicmatter and soil mineral components to obtain organo-mineralcomplexes with a range of organic C loadings. The relative biolog-ical stability of the C in the organo-mineral complexes generatedfrom DOM sorption was subsequently assessed using thermalanalysis and laboratory incubation experiments.

2. Materials and methods

The current study was constructed as a two-step experimentaldesign. The first step consisted of several batch sorption experi-ments designed to create organo-mineral complexes with differingorganic C loadings and to determine whether these complexesexhibit organic C saturation behavior. The second step consisted ofthermal analysis and laboratory incubations to assess the relativestability of the C sorbed on the organo-mineral complexes.

2.1. Batch sorption experiment

The selection of the mineral and organic materials used in thebatch sorption experiments was made with the goal of simulatingphenomena observed in the field as directly as possible, whileincreasing the likelihood of observing C saturation. For this reason,soil samples from sub-surface horizons were selected instead ofneat and pure minerals present in soils (e.g., kaolinite, illite,

goethite, gibbsite, etc.), recognizing that soil samples contain asignificant amount of heterogeneity in mineral composition. Sub-surface soils were selected rather than surface soils because oflow initial C concentrations that enable them to sorb greateramounts of organic C inputs as well as avoiding issues of dis-tinguishing between newly sorbed versus existing and potentiallydesorbed organic matter.

Four soils (referred to here by their soil series names: Edgemont,Drummer, San Ysidro, and Towaliga) were selected for use in thebatch sorption experiments (Table 1). The Edgemont soil wascollected from the Stroud Water Research Center, Avondale PA.Three other soils were selected from a set of 213 subsurface sam-ples used in a previous study on DOM sorption (Mayes et al., 2012).The selected soils represent three soil orders, and were chosen onthe basis of having a maximum DOC sorption capacity (Qmax) in theupper 25th percentile of their respective soil order based on ex-periments conducted by Mayes et al. The soils contained someinitial organic matter, but C concentrations were <0.5% (Table 1).While clay and Fe oxide contents were found to be the strongestpredictors of Qmax (Kothawala et al., 2009; Mayes et al., 2012), theselected samples used in the current study varied in Fe oxidecontent and mineralogy, but did not vary widely in clay content(Table 1). Based on Natural Resources Conservation Service data forthe soil series (http://soils.usda.gov/technical/classification/osd/index.html), mineralogy of the Edgemont soil is mixed, character-ized by kaolinite as the dominant mineral in the clay fraction, withdetectable amounts of illite, vermiculite, montmorillonite, chloriteand interstratified clays. The Bt horizon where samples werecollected from is typified as containing more than 40% silt. TheDrummer soil is characterized by smectite as the dominant claymineral in the upper part of the series control section and illite inthe lower part. The sample contained 32% clay and 53% silt. Thedominant clay mineral in the San Ysidro soil is smectite, and the Bhorizon where samples were collected contained 33% clay and 31%silt. The dominant clay mineral in the Towaliga soil is kaolinite, andthe samples contained 33% clay and 24% silt. (Textural data for thelatter three soils was obtained byMayes, personal communication.)

Soils were ground to pass through a 500-mm sieve and storedair-dry prior to use in the batch sorption experiments. Soil mineralspecific surface area (SSA) wasmeasured on three replicate aliquotsof soil prior to the batch sorption experiments using the N2-BETmethod after pre-treating samples to remove organic matter.Organic matter removal was performed by placing samples in amuffle furnace at 350 �C for 18 h Wagai et al. (2009) previouslydemonstrated greater organic matter removal by muffling than byhypochlorite treatment, but expressed caution due to the potentialalteration of iron oxides. Our samples contained sufficiently lowconcentrations of iron oxides, that we did not consider this to be asubstantial problem, but as Wagai et al. concluded, the measuredvalues should be consider minimum values. Samples were also de-gassed at 325 �C for 4 h with N2 and He to remove adsorbed water.Nitrogen was dosed on the surfaces at 77 K with different gaspressures in a Tristar 300 surface area and porosity analyzer(Micromeritics, Norcross, GA). The multi-point Brunauer-Emmett-Teller (BET) method was used to calculate SSA values under therelative pressure between 0.05 and 0.3 atm.

Similar to the approach used to select the mineral component, ahigh C concentration, chemically complex DOM solution derivedfrom leaf litter was selected instead of a commercial humic acid orsingle, model compounds. A stock DOM solution was prepared bymixing surface leaf litter with water (10:1 water volume to littermass ratio) in a 40 L plastic carboy. The litter was collected from amixed oak-maple stand, with dominant species consisting ofQuercus palustris, Quercus rubra, Quercus alba, Acer rubrum, Lir-iodendron tulipifera, and Fagus grandifolia at the Stroud Water

Table 1Characteristic properties of four soils used for dissolved organic matter batch sorption experiments. Initial C concentrations were measured prior to the batch sorption ex-periments, and Final C concentrations were the solid-phase organic C concentrations after batch sorptionwith the highest dissolved organic matter concentration used in theexperiment.

Site Soil taxonomy Depth (cm) Silt þ clay (%) pHH2O Total Fe (g kg�1) Mineral specificsurface area (SSA, m2 g�1)

Initial C (mg g�1) Final C (mg g�1)

Edgemont Hapludults 54e67 56.6 6.26 23.8 24.9 2.81 � 0.15 28.1 � 1.1Drummer Endoaquolls 50e75 85.0 7.51 45.0 35.3 5.78 � 0.14 30.4 � 0.1San Ysidro Palexeralfs 36e76 64.2 6.99 11.4 33.2 4.96 � 0.02 24.8 � 0.6Towaliga Hapludults 10e21 57.4 7.70 21.9 23.5 3.28 � 0.22 25.1 � 1.1

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405400

Research Center, in the same area from where the Edgemont soilsample was collected. The litter was kept field moist, cut into smallpieces (1e3 cm), added to the water, and allowed to soak for 48 h atroom temperature with occasional stirring to invert the mixture.The resulting suspension was decanted, passed through glass fiberfilter (GF/F, 0.45 mm diameter), and the DOM filtrate was stored at4 �C until subsequent use. The resulting stock DOM solution wasfound to have a dissolved organic carbon (DOC) concentration of614 � 7.6 mg C L�1 a C/N ratio of 21.5 and isotopic signaturesof �28.7& d13C and 0.5& d15N.

The batch sorption experiments were conducted by mixing0.25 g aliquots of soil with 100 mL of DOM in a 250 mL Erlenmeyerflask. Solutions differing in DOM concentrations were prepared bydiluting the stock DOM solutionwith deionizedwater to target DOCconcentrations of 50, 100, 150, 200, 300, 350, 400, 450, 500, and600 mg C L�1. The batch sorption experiment for each soil includedone control sample consisting of DOM (w600 mg C L�1) withoutsoil to determine DOM degradation during the experiment, and asecond control composed of 0.25 g soil mixed with deionized water(0 mg C L�1). Thus, a total of 12 Erlenmeyer flasks for each exper-iment were shaken for 24 h on a reciprocal shaker (180 rpm) placedinside a constant temperature chamber set to 20 (�0.2) �C. Tomaximize DOM sorption, we did not use a biocide and performedthe experiments at room temperature instead of the more typicaltemperature of 4 �C. Previous experiments showed that microbialactivity enhanced sorption (A. K. Aufdenkampe, unpublishedwork). After 24 h, the suspensions were passed through a 0.45 mmmembrane filter (GE Osmonics, Minnetonka, MN). The organo-mineral complexes retained on the filters were washed intoaluminum pans and oven-dried at 50 �C. The filtrates were stored at4 �C in glass vials for amaximum of 24 h before DOC determination.Batch sorption experiments were replicated three times for eachsoil, and each replicate set of experiments was run independently.

The original stock DOM solution, the initial dilution series andthe filtrates after the batch sorption experiments were analyzed forDOC using a persulfate digestion and spectrophotometric method(Method 10173, Hach Company, Loveland CO). Soild-phase carbonand nitrogen concentrations, and d13C and d15N isotopic signaturesof the organo-mineral complexes retrieved after DOM sorptionwere analyzed using an elemental analyzer (Costech AnalyticalTechnologies, Valencia, CA) coupled to a Deltaþ stable isotope ratiomass spectrometer (Finnigan MAT, Bremen, Germany). Organic Cloadings of the organo-mineral complexes were then determinedby dividing the solid-phase C concentrations (mg C g�1 sample)after DOM sorption by the mineral SSA (m2 g�1 sample) valuesdetermined on the initial soil aliquots.

2.2. Thermal analysis

The stability of sorbed organic matter was characterized usingevolved CO2 gas analysis (CO2-EGA) during thermal analyses. Ali-quots from a subset of samples generated in the DOM sorptionexperiments were analyzed using a Netzsch simultaneous heat flux

thermal analyzer (STA 409 PC Luxx) coupled to an LI-840 CO2/H2Oinfrared gas analyzer (IRGA; LI-COR Biosciences, Lincoln NE).Samples were heated from ambient (w25 �C) to 105 �C at10 �C min�1, held at this temperature for 15 min to remove samplemoisture, then heated to 800 �C at 10 �C min�1. CO2 concentrationsin the evolved gas were recorded as partial pressure (ppm CO2), andthe temperature at which one-half of the CO2 had evolved (CO2-T50)was determined as an index of the thermal stability of the sorbedorganic matter (Fernández et al., 2012), with the underlyingassumption that thermal stability is related to biological stability(Plante et al., 2011).

2.3. Laboratory incubation

A laboratory incubation experiment was used to assess therelative biological stability of the organic matter associated withthe organo-mineral complexes generated in the DOM sorptionexperiments. Approximately 0.2 g of organo-mineral complexeswas mixed with 0.2 g of acid-washed sand in 4 mL Exetainerglass vials (Labco Limited, UK), and mixed for 10 s on a vortexshaker. The sand was added to increase the porosity of the incu-bation matrix and improve aeration. Deionized water was added tothe glass vials to 55% water-filled pore space to ensure optimalmoisture for microbial activity. Vials were kept openwith the septacaps placed loosely on the vials, and placed in a glass desiccatorwith the bottom filled with water to maintain high humidity.Samples were incubated for 60 days at 25 (�0.2) �C. Respirationrates were determined at incubation days 1, 4, 7, 10, 15, 20, 30, 40,50 and 60. Prior to CO2 analysis, each vial was capped and flushedwith 20 mL CO2-free air, then returned to the incubator for 24 hearly in the incubation experiment and for 48 h later in the incu-bation, to allow sufficient CO2 to accumulate in the headspace ofthe vial. The small headspace volume in the vials increased thepotential for measurable changes in CO2 concentration generatedby the small mass of samples available, but required a small samplealiquot volume for analysis. CO2 concentrations were determinedby gas chromatography-mass spectrometry (GCeMS). Ninety mL ofheadspace sample were injected into a Trace GC (Thermo Scientific,Asheville, NC) coupled to a Delta þ stable isotope mass spectrom-eter (Finnigan MAT, Bremen, Germany) via an open split. Carbondioxide concentrations were calculated by calibration of peak areas,and isotope data were not used.

2.4. Data analysis

The amount of DOM sorbed onto the soils to generate organo-mineral complexes with differing organic C loadings was deter-mined in three ways. The first method was to calculate the differ-ence in DOC concentrations of the solution phase before and afterthe batch sorption experiment, without accounting for C losses dueto decomposition (referred to as DDOC). The second method was tocalculate corrected differences in DOC concentrations before andafter sorption by subtracting C losses attributable to decomposition

Fig. 1. Soil C loading (mg C m�2), soil C/N, and stable isotopic signature (d13C and d15N,&) of organo-mineral complexes after 24 h batch sorption of four soils (Edgemont, Drummer,San Ysidro, and Towaliga) with differing initial dissolved organic C concentrations (mg C L�1). Vertical line separates blank samples treated with water only from samples treatedwith dissolved organic matter.

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405 401

(referred to as DDOCcorr). Decomposition C losses were estimatedfrom decreases in the DOC concentration of the soil-free blanksample during each set of 24-h batch experiments. The thirdmethod was to calculate the difference in solid phase (soil) C con-centrations before and after the DOM sorption experiment(referred to as DSOC).

The Langmuir isotherm is frequently used to relate the amountof a species adsorbed on a solid surface to the solution concentra-tion. We compared the amounts of DOM sorption on soils as afunction of initial C concentrations of DOM to linear and Langmuirregression models to test whether or not the organo-mineralcomplexes were potentially exhibiting C saturation behavior.Model fitting comparisons were performed using Akaike’s Infor-mation Criterion with CurveExpert Professional v.1.5 (Daniel G.Hyams, www.curveexpert.net). In addition, current conceptualthinking suggests that a threshold C loading may exist at whichstabilized, mineral-bound SOM becomes more labile because ofweaker binding. The threshold soil C loading has been yet beendefinitively identified, but a value of 1 mg Cm�2 has been proposedas a potential maximal amount of organic C that can be stabilizedon mineral particles (Feng et al., 2013; Mayer, 1994). Carbon load-ings greater than 1 mg C m�2 have previously been observed, butare generally found in low density fractions that may include asignificant contribution from POM (Mayer and Xing, 2001; Wagaiet al., 2009). The organo-mineral complexes generated in the cur-rent study are predominantly mineral with coatings of sorbedDOM, and thus the so-called monolayer equivalent (w1 mg C m�2)may be an appropriate maximum. We make no assumptions aboutthat spatial distribution of organic matter on mineral surfaces (e.g.,monolayer), only that this C loading value may represent athreshold beyond strongly and weaklymineral-bound SOM. For the

purposes of this study, we defined soil C saturation behavior as theobservation of a pattern of DOM sorption with increasing initialDOM fitting an asymptotic regression, or as a maximum organic Cloading that approaches, but does not substantially exceed1 mg C m�2 with incrasing initial DOC concentration.

Cumulative CO2 respired during the 60-day incubation wasdetermined using respiration rates measured periodically duringthe incubation. Respiration rates were calculated by subtractingrates from control samples (i.e., samples that went through thebatch sorption experiment with 0 mg C L�1) from treated samples,then normalizing by the amount of sorbed C (i.e., g CO2eC (100 gsorbed-C)�1), which was determined using the solid-phase (DSOC)method. Relative biological stability indices were determined byfitting cumulative CO2 respiration data to the first-order expo-nential equation Ct ¼ Co (1 � e�kt), where Ct is cumulative CO2respiration, Co is the potentially mineralizable C, and k is the first-order decomposition constant. Data from each of the three incu-bated replicates from each soil-DOM concentration combinationwere fit individually using non-linear regression (Sigmaplot 11.0,Systat Software Inc.).

3. Results and discussion

3.1. DOM sorption

Measured values of C/N, d13C, and d15N of organo-mineralcomplexes generated by DOM sorption were significantly smalleror more negative (P < 0.01) than values of initial soil materialssubjected only to water (Fig. 1), and also differed from the values ofthe initial stock DOM. Values of d15N and C/N of organo-mineralcomplexes obtained from DOM sorption were smaller compared

Fig. 3. Temperature at which half of the evolved CO2 is generated during thermalanalysis (CO2-T50) as a function of organic C loading on organo-mineral complexesfrom four soils (Edgemont, Drummer, San Ysidro, and Towaliga).

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405402

to d15N values and C/N ratios of each type of four stock soils used inthe sorption experiments (Fig. 1). C/N ratios of the organo-mineralcomplexes remained relatively constant as a function of initial DOCconcentration (and therefore of final C loading), where P-values forregression slopes were generally non-significant. However, d13Cand d15N values generally decreased as C loading increased, withthe exception of San Ysidro, which showed no statistically signifi-cant trends for either d13C or d15N. While changes in C/N of theorgano-mineral complexes would have been a more substantialindication of changes in the composition of the sorbed organicmatter, consistent with the concept of molecular fractionationduring DOM sorption (Kaiser et al., 2001; Oren and Chefetz, 2012),it is unlikely that the changes in isotopic signature can be attributedto isotopic fractionation alone. We therefore speculate that thecomposition of the organic matter in the organo-mineral com-plexes changed as the amount of sorbed organic matter increased.Another, likely minor, mechanism for changes in compositionwould be through exchange of organic matter during desorption-sorption of native versus added organic matter.

The amount of DOM sorbed onto the four soils differed signifi-cantly based on how it was calculated, particularly at greater initialDOM concentrations (Fig. 2). Negative amounts of C sorptionobserved at initial DOM concentrations less than 100 mg C L�1

indicate that C from the original soil material was being desorbed.At larger initial DOM concentrations, DDOC and DDOCcorr estimatesof C sorption were significantly greater than DSOC estimates. Thesedifferences may be because DOM decomposition rates were rela-tively slow when initial DOM concentrations were small, andincreased as initial DOM concentrations increased. Larger estimatesof DOM sorption by the DDOC and DDOCcorr methods compared tothe DSOC method may also be attributed to enhanced decompo-sition in the presence of the mineral phase, which was not sterile,remembering that the correction for DOM decompositionwas doneusing a soil-free blank. The results demonstrate that the DDOCmethods commonly used to determine organic C sorption (e.g.,Schneider et al., 2010) may overestimate the actual amount oforganic C sorbed on soils. The amount of C sorbed based on thesolid phase (DSOC) measurements is used in our subsequent ana-lyses and discussion because it is a direct measurement of C in the

Fig. 2. Comparison of amounts of sorbed dissolved organic matter in four soils(Edgemont, Drummer, San Ysidro, and Towaliga). Three methods to calculate DOMsorption are: direct measurement of soil C concentration after sorption experiment(DSOC, C), difference of C concentrations of DOM before and after sorption experi-ment corrected for decomposition (DDOCcorr, 6), and uncorrected difference of Cconcentrations of DOM before and after sorption experiment (DDOC, :). Data aremean � SE (n ¼ 3).

organo-mineral complexes generated, and therefore judged to be abetter estimate of the amount of C being assessed for stability in theincubation experiment.

The amount of C sorbed on the soils increased with increasinginitial DOM concentrations, up to 30.4 � 0.1 mg C g�1 in theDrummer soil, 28.1 �1.1 mg C g�1 in Edgemont, 25.1 �1.1 mg C g�1

in Towaliga, and 24.8 � 0.6 mg g�1 in San Ysidro (Fig. 2). Thesesorbed C concentrations were substantially greater than the Qmaxestimates generated by Mayes et al. (2012), which is attributable toboth the method used for determining the amount of sorbed C anddiffering experimental conditions during the sorption experiment.For instance, the largest C to soil ratio used in the sorption exper-iments of Mayes et al. (2012) was 6 mg C g�1 soil, versus24 mg C g�1 soil in the current study. The differences betweenstudies are relatively conservative because the y are based on thesmaller estimates of sorbed C generated by the solid phase (DSOC)method. Amounts of C sorbed showed linear relationships withinitial DOC concentration for all soils, and did not exhibit the samedegree of asymptotic behavior as previously observed by Mayeset al. (2012). Model comparisons indicated that the likelihoodthat the better model was the linear model was >70% for each soil.The asymptotic relationships frequently observed in sorption ex-periments may be due to relatively low diffusion gradients gener-ated by low DOC concentrations (<100 mg C L�1 in Mayes et al.,2012), while the current study used much greater DOC concen-trations (up to 600 mg C L�1).

The increasing C concentrations of the organo-mineral com-plexes after sorption also resulted in increased C loadings.Maximum carbon loadings were 0.75 mg C m�2 in the San Ysidrosoil, 0.86 mg C m�2 in Drummer, 1.07 mg C m�2 in Towaliga, and1.19 mg C m�2 in Edgemont. Carbon loadings calculated using theQmax estimates from Mayes et al. (2012) were significantly lowerthan our observations, ranging from 0.08 to 0.18 mg C m�2. Anorganic C loading of 1 mg C m�2 has previously been proposed as amaximal C loading (Feng et al., 2013; Mayer, 1994). The highest Cloadings of the San Ysidro and Drummer soils were less than1 mg C m�2, suggesting that these two soils had not reached Csaturation. The highest C loadings of the Towaliga and Edgemontsoils were greater than 1 mg C m�2, though not by a large margin.Overall, the observed linear relationships and the lack of C loadingsmuch greater than 1 mg C m�2 suggest that these soils may nothave reached C saturation under the conditions of our batch

Fig. 4. Patterns of CO2 evolution during thermal analysis of organo-mineral complexesfrom the Edgemont soil.

Fig. 5. Total (mg CO2eC kg�1 soil, left column) and sorbed C normalized (g CO2eC 100 g�1 soorgano-mineral complexes from four soils (Edgemont, Drummer, San Ysidro, and Towaliga).

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405 403

sorption experiments. While there was little evidence for C satu-ration behavior, the organic C loadings of the organo-mineralcomplexes seemed to span a sufficiently wide range (0.2e1.2 mg Cm�2) to test for differences in SOM stability as a function ofC loading.

3.2. Stability of sorbed C in organo-mineral complexes

Results of CO2-EGA analyses generally did not support the hy-pothesis that sorbed C stability decreases with increasing C loading.Values of CO2-T50 generally increased with increasing C loading forall four soils (Fig. 3), though the regressions for the Drummer andSan Ysidro soils were not statistically significant due to a combi-nation of large scatter and small slope. Greater CO2-T50 valuesindicate a greater energy input required to combust the sorbedorganic matter, which may be attributable to either a shift ingreater chemical complexity as noted above in reference to shifts inisotopic signature, or to an increase in the sorption bond strength.The overall shape of the CO2-EGA thermograms did not showsubstantial differences with C loading that would be indicative ofmajor changes in chemical composition (Fig. 4), though increasing

rbed-C, right column) cumulative CO2 respired during a 60-day laboratory incubation ofOpen symbols are control soils that did not undergo dissolved organic matter sorption.

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405404

C loadings resulted in small decreases in thermally labile materialcombusted atw275 �C, as indicated by the diminishing shoulder onthe main peak of CO2 evolution, and small accumulations of morethermally stable organic matter combusted at w400e500 �C, asindicated by the growing shoulder on the main peak of CO2evolution.

Total cumulative CO2 respiration (mg CO2eC kg�1 soil) over the60-day incubation increased with increasing C loading for all foursoils (data not shown). However, when normalized by the C con-centration of the organo-mineral complexes (SOCf; Fig. 5, left col-umn), CO2 respiration decreased with increasing C loadingsinitially, then reached a relatively stable value. This pattern wasmore substantial when respiration data were normalized by theamount of C sorbed (DSOC; Fig. 5, right column), where initialnormalized respiration were much higher initially.

Results of exponential model fitting found that estimates of theproportion of potentially mineralizable C (i.e., C0) were similar tomeasured values of C-normalized cumulative CO2 respired, sug-gesting that the 60-day incubation was sufficiently long to allowthe respiration of the easily mineralizable organic matter pool in allfour soils (Table S1). The decreasing C0 values with increasing Cloading indicate that the relative size of biologically labile organicmatter in the organo-mineral complexes decreased as organic Cloading increased. Modeled decomposition rate constants (i.e., k)initially increased with increasing soil C loadings, then appeared tolevel off (Fig. 6). This pattern was less pronounced in the Drummersoil, likely due to large scatter (comparable to Edgemont) and thelack of high C loadings (compared to Edgemont and Towaliga). In-creases in the decomposition constant indicate that the mineral-izable pool of sorbed C is being degraded more rapidly. Theopposing trends of normalized C0 and of kwith increasing C loadingsuggest that the size and stability of biologically labile organicmatter sorbed on soil minerals may be independent of each other.

These results differ from the prediction of soil C saturationbehavior that SOM stability decreases as organic C inputs andstocks increase. Studies in which soil C saturation behavior hasbeen observed (e.g., Chung et al., 2010, 2008; Gulde et al., 2008)were based on changes in C concentrations in organo-mineralcomplexes isolated from field soils subjected to long-term agro-nomic C inputs. Decreased bulk SOC storage efficiency of bulk soilwith increasing C inputs/C stocks was likely caused by increased

Fig. 6. Soil organic C decomposition constants (day�1) from models of CO2 respirationduring 60-day laboratory incubation of organo-mineral complexes from four soils(Edgemont, Drummer, San Ysidro, and Towaliga).

contributions of POM rather than increased decomposition oforganic matter stabilized on soil minerals. No assessment ofchanges in SOM stability in organo-mineral complexes was per-formed in previous studies. Conversely, SOM in field soils is sub-jected to a range of stabilization mechanisms (e.g., physicalprotection by aggregation) that did not occur in our simplifiedstudy. In the current study, organic matter stability was assessedusing rapidly formed, non-aged organo-mineral complexes. Whilethe complexes were generated from complex end-member mate-rials (e.g., mineral subsoils and litter extracts) rather than modelcompounds and minerals, the anticipated changes in stability oforganic matter with increasing C loading may only be observable inthe long-term.

Very high DOM concentrations were used in the sorption ex-periments to achieve the high C loadings observed. At these highconcentrations, the dissolved organic molecules may have hadgreater probabilities of interacting with other organic molecules.The C-normalized respiration results suggest that the organicmolecules at higher C loadings may be associated with one anothervia weak forces such as hydrophobic interactions to form co-agulates rather than bound tomineral surfaces. So it seems possiblethat microbes may have had more limited accessibility to the in-teriors of such organic coagulates compared to the organic mattersparsely associated with mineral surfaces in the samples withlower C loadings. In addition, while the modeled C0 values matchedthe measured values at the end of the 60-day incubation, theproportion of sorbed C that was respired was quite small because ofthe short duration of the incubation experiments. Bottcher (2004)demonstrated that parameters of first-order mineralizationmodels can be strongly affected by incubation length. If the in-cubations were conducted for much longer durations, it may bepossible to find an opposite trend (i.e., increase in the C-normalizedrespiration with increasing C loading). While microbes may beunable to access most of the sorbed organic matter at high Cloadings in 60 days, they may be able to utilize greater proportions(compared to the samples with smaller C loadings) in time due tothe weak binding forces among organic molecules for the reasonsmentioned above.

4. Conclusion

Thermal and biological indices of SOM stability in organo-mineral complexes generated from batch DOM sorption experi-ments generally increased as a function of C loading. These resultsdo not support our hypothesis that SOM stability decreases withincreasing C loading. While the conditions of the DOM sorptionexperiments achieved significantly larger C loadings than previousstudies, the amount of C sorbed appears to have been insufficient toinduce C saturation behavior and the alteration of organo-mineralinteractions from strong bindings to weaker bindings, as sug-gested by the zonal model (Kleber et al., 2007). It is possible thatorgano-mineral associations are relatively strong and consistentuntil the threshold of C saturation is reached, and not until then aredecreases in organic matter stability observable. In addition, themaximal C loading of organo-mineral complexes at C saturationmight be greater than 1mg Cm�2. Hence, further work is needed toseek longer-lived organo-mineral complexes (either natural orlaboratory-generated) that exhibit C saturation to better test forchanges in SOM stability as a function of increasing C inputs.

Acknowledgements

The authors wish to thank an anonymous reviewer forconstructive comments concerning the interpretation of the labo-ratory incubation results, andMelanie Mayes at ORNL for supplying

W. Feng et al. / Soil Biology & Biochemistry 69 (2014) 398e405 405

the Drummer, San Ysidro and Towaliga soil samples. Funding wasprovided by DOE-BER grant # ER63912.

Appendix A. Supplementary data

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.soilbio.2013.11.024

References

Arnarson, T.S., Keil, R.G., 2000. Mechanisms of pore water organic matter adsorptionto montmorillonite. Mar. Chem. 71, 309e320.

Balesdent, J., Mariotti, A., Guillet, B., 1987. Natural 13C abundance as a tracer forstudies of soil organic matter dynamics. Soil Biol. Biochem. 19, 25e30.

Balesdent, J., Wagner, G.H., Mariotti, A., 1988. Soil organic-matter turnover in long-term field experiments as revealed by C-13 natural abundance. Soil Sci. Soc. Am.J. 52, 118e124.

Bottcher, J., 2004. Uncertainties of nonlinearly estimated parameters from in-cubations of soil organic matter. J. Plant Nutr. Soil Sci. Zeitschrift Fur Pflanze-nernahrung Und Bodenkunde 167, 293e302.

Campbell, C.A., Bowren, K.E., Schnitzer, M., Zentner, R.P., Townleysmith, L., 1991.Effect of crop rotations and fertilization on soil organic matter and somebiochemical properties of a thick black chernozem. Can. J. Soil Sci. 71, 377e387.

Christensen, B.T., 1998. Carbon in primary and secondary organo-mineral com-plexes. In: Carter, M.R., Stewart, B.A. (Eds.), Structure and Organic MatterStorage in Agricultural Soils. CRC Press, Boca Raton, pp. 97e165.

Chung, H., Ngo, K.J., Plante, A.F., Six, J., 2010. Evidence for carbon saturation in ahighly structured and organic matter-rich soil. Soil Sci. Soc. Am. J. 74, 130e138.

Chung, H.G., Grove, J.H., Six, J., 2008. Indications for soil carbon saturation in atemperate agroecosystem. Soil Sci. Soc. Am. J. 72, 1132e1139.

Feng, W., Plante, A.F., Six, J., 2013. Improving estimates of maximal organic carbonstabilization by fine soil particles. Biogeochemistry 112, 81e93.

Fernández, J.M., Peltre, C., Craine, J.M., Plante, A.F., 2012. Improved characterizationof soil organic matter by thermal analysis using CO2/H2O evolved gas analysis.Environ. Sci. Technol. 46, 8921e8927.

Gu, B.H., Schmitt, J., Chen, Z.H., Liang, L.Y., McCarthy, J.F., 1994. Adsorption anddesorption of natural organic matter on iron oxide: mechanisms and models.Environ. Sci. Technol. 28, 38e46.

Gulde, S., Chung, H., Amelung, W., Chang, C., Six, J., 2008. Soil carbon saturationcontrols labile and stable carbon pool dynamics. Soil Sci. Soc. Am. J. 72, 605e612.

Hassink, J., 1996. Preservation of plant residues in soils differing in unsaturatedprotective capacity. Soil Sci. Soc. Am. J. 60, 487e491.

Huggins, D.R., Buyanovsky, G.A., Wagner, G.H., Brown, J.R., Darmody, R.G., Peck, T.R.,Lesoing, G.W., Vanotti, M.B., Bundy, L.G., 1998. Soil organic C in the tallgrassprairie-derived region of the corn belt: effects of long-term crop management.Soil Tillage Res. 47, 219e234.

Kahle, M., Kleber, M., Jahn, R., 2002. Carbon storage in loess derived surface soilsfrom Central Germany: Influence of mineral phase variables. J. Plant Nutr. SoilSci. Zeitschrift Fur Pflanzenernahrung Und Bodenkunde 165, 141e149.

Kaiser, K., Guggenberger, G., Zech, W., 2001. Isotopic fractionation of dissolvedorganic carbon in shallow forest soils as affected by sorption. Eur. J. Soil Sci. 52,585e597.

Kleber, M., Sollins, P., Sutton, R., 2007. A conceptual model of organo-mineral in-teractions in soils: self-assembly of organic molecular fragments into zonalstructures on mineral surfaces. Biogeochemistry 85, 9e24.

Kothawala, D.N., Moore, T.R., Hendershot, W.H., 2009. Soil properties controlling theadsorption of dissolved organic carbon to mineral soils. Soil Sci. Soc. Am. J. 73,1831e1842.

Mayer, L.M., 1994. Relationships between mineral surfaces and organic carbonconcentrations in soils and sediments. Chem. Geol. 114, 347e363.

Mayer, L.M., Xing, B.S., 2001. Organic matter-surface area relationships in acid soils.Soil Sci. Soc. Am. J. 65, 250e258.

Mayes, M.A., Heal, K.R., Brandt, C.C., Phillips, J.R., Jardine, P.M., 2012. Relation be-tween soil order and sorption of dissolved organic carbon in temperate subsoils.Soil Sci. Soc. Am. J. 76, 1027e1037.

Oren, A., Chefetz, B., 2012. Sorptive and desorptive fractionation of dissolvedorganic matter by mineral soil matrices. J. Environ. Qual. 41, 526e533.

Paustian, K., Collins, H.P., Paul, E.A., 1997. Management controls on soil carbon. In:Paul, E.A., Paustian, K., Elliott, E.T., Cole, C.V. (Eds.), Soil Organic Matter inTemperate Agroecosystems. CRC Press, Boca Raton, FL, pp. 15e49.

Plante, A.F., Fernández, J.M., Haddix, M.L., Steinweg, J.M., Conant, R.T., 2011. Bio-logical, chemical and thermal indices of soil organic matter stability in fourgrassland soils. Soil Biol. Biochem. 43, 1051e1058.

Schneider, M.P.W., Scheel, T., Mikutta, R., van Hees, P., Kaiser, K., Kalbitz, K., 2010.Sorptive stabilization of organic matter by amorphous Al hydroxide. Geochim.Cosmochim. Acta 74, 1606e1619.

Six, J., Conant, R.T., Paul, E.A., Paustian, K., 2002. Stabilization mechanisms of soilorganic matter: Implications for C-saturation of soils. Plant Soil 241, 155e176.

Sollins, P., Kramer, M.G., Swanston, C., Lajtha, K., Filley, T., Aufdenkampe, A.,Wagai, R., Bowden, R., 2009. Sequential density fractionation across soils ofcontrasting mineralogy: evidence for both microbial- and mineral-controlledsoil organic matter stabilization. Biogeochemistry 96, 209e231.

Soon, Y.K., 1998. Crop residue and fertilizer management effects on some biologicaland chemical properties of a Dark Grey Solod. Can. J. Soil Sci. 78, 707e713.

Stevenson, F.J., 1994. Humus Chemistry. Genesis, Composition, Reactions. JohnWiley & Sons, New York, NY.

Stewart, C.E., Paustian, K., Conant, R.T., Plante, A.F., Six, J., 2007. Soil carbon satu-ration: concept, evidence and evaluation. Biogeochemistry 86, 19e31.

Stewart, C.E., Paustian, K., Conant, R.T., Plante, A.F., Six, J., 2008a. Soil carbon satu-ration: evaluation and corroboration by long-term incubations. Soil Biol. Bio-chem. 40, 1741e1750.

Stewart, C.E., Plante, A.F., Paustian, K., Conant, R.T., Six, J., 2008b. Soil carbon satu-ration: linking concept and measurable carbon pools. Soil Sci. Soc. Am. J. 72,379e392.

Trumbore, S., 2000. Age of soil organic matter and soil respiration: Radiocarbonconstraints on belowground C dynamics. Ecol. Appl. 10, 399e411.

Wagai, R., Mayer, L.M., Kitayama, K., 2009. Extent and nature of organic coverage ofsoil mineral surfaces assessed by a gas sorption approach. Geoderma 149, 152e160.

West, T.O., Six, J., 2007. Considering the influence of sequestration duration andcarbon saturation on estimates of soil carbon capacity. Clim. Change 80, 25e41.