Testing for soil carbon saturation behavior in agricultural soils receiving long-term manure amendments

Transcription

1 Testing for soil carbon saturation behavior in agricultural soils receiving long-term manure amendments W. Feng 1,M. Xu 2, M. Fan 3, S. S. Malhi 4, J. J. Schoenau 5, J. Six 6, and A. F. Plante 1 1 Department of Earth and Environmental Science, University of Pennsylvania, Philadelphia, PA , USA; 2 Ministry of Agriculture Key Laboratory of Crop Nutrition and Fertilization, Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing , China; 3 College of Resources and Environmental Sciences, China Agricultural University, Beijing 10094, China; 4 Agriculture and Agri-Food Canada, Melfort, Saskatchewan, Canada S0E 1A0; 5 Department of Soil Science, University of Saskatchewan, Saskatoon, Saskatchewan, Canada S7N 5A8; and 6 ETH Zurich, 8092 Zurich, Switzerland. Received 28 February 2013, accepted 15 November Published on the web 21 November Feng, W., Xu, M., Fan, M., Malhi, S. S., Schoenau, J. J., Six, J. and Plante, A. F Testing for soil carbon saturation behavior in agricultural soils receiving long-term manure amendments. Can. J. Soil Sci. 94: Agricultural soils are typically depleted in soil organic matter compared with their undisturbed counterparts, thus reducing their fertility. Organic amendments, particularly manures, provide the opportunity to restore soil organic matter stocks, improve soil fertility and potentially sequester atmospheric carbon (C). The application of the soil C saturation theory can help identify soils with large C storage potentials. The goal of this study was to test whether soil C saturation can be observed in various soil types in agricultural ecosystems receiving long-term manure amendments. Seven long-term agricultural field experiments from China and Canada were selected for this study. Manure amendments increased C concentrations in bulk soil, particulate organic mattersand, and siltclay fractions in all the experiments. The increase in C concentrations of siltclay did not fit the asymptotic regression as a function of C inputs better than the linear regression, indicating that siltclay did not exhibit C saturation behavior. However, 44% of calculated C loading values for siltclay were greater than the presumed maximal C loading, suggesting that this maximum may be greater than 1 mg C m 2 for many soils. The influences of soil mineral surface properties on C concentrations of siltclay fractions were site specific. Fine soil particles did not exhibit C saturation behavior likely because current C inputs were insufficient to fill the large C saturation deficits of intensely cultivated soils, suggesting these soils may continue to act as sinks for atmospheric C. Key words: Fertilization, long-term, manure, siltclay, soil carbon saturation Feng, W., Xu, M., Fan, M., Malhi, S. S., Schoenau, J. J., Six, J. et Plante, A. F E tude de la saturation en carbone dans les sols agricoles longtemps bonifie s avec du fumier. Can. J. Soil Sci. 94: Habituellement, les sols agricoles s ave` rent plus pauvres en matie` re organique que les sols similaires laisse s intacts, ce qui en re duit la fertilite. Les amendements organiques le fumier, en particulier renouvellent les re serves de matie` re organique, rendent le sol plus fertile et, éventuellement, séquestrent le carbone atmosphe rique (C). L application de la théorie de la saturation du sol en C permet d e tablir quels sols pourraient stocker une grande quantite de C. La pre sente étude devait ve rifier si cette saturation s observe dans les sols de divers type des e cosyste` mes agricoles amendés pendant longtemps avec du fumier. A` cette fin, les auteurs ont examine le sol de parcelles agricoles de Chine et du Canada issu de sept expe riences de longue haleine. Dans tous les cas, l application de fumier augmente la concentration de C dans le sol brut, dans la fraction matie` re organique particulaire/sable et dans la fraction limon/argile. La re gression asymptotique reposant sur les apports de C n explique pas mieux la hausse de la concentration de C dans la fraction limon/argile que la régression linéaire, signe que cette fraction n illustre pas le comportement de saturation en C. Ne anmoins, 44 % des valeurs calcule es pour la charge de C dans la fraction limon/argile de passaient la charge maximale pre sume e. On en de duit que cette valeur maximale pourrait être supe rieure a` 1 mg de C par m 2 dans de nombreux sols. L influence des proprie te s de la surface mine rale du sol sur la concentration de C dans la fraction limon/argile de pend du site. Les fines particules de sol n illustrent pas le comportement de saturation en C, sans doute parce que les apports en C courants ne compensent pas la forte carence au niveau de la saturation en C observée dans les sols intense ment cultivés, ce qui pourrait vouloir dire que les sols de ce type continuent de servir de puits au carbone pre sent dans l atmosphe` re. Mots clés: Fertilisation, long terme, fumier, limon/argile, saturation en carbone du sol Organic carbon (C) stored in croplands ( Pg C) accounts for approximately 10% of total soil C up to 1 m deep (1500 Pg C) globally (Eswaran et al. 1993; Paustian et al. 1997; Jobbagy and Jackson 2000). Soils in croplands are frequently C depleted compared with soil C levels before cultivation, which enables cropland soils to serve as a potential C sink to offset the increasing rate of atmospheric CO 2 increase due to fossil fuel emissions. It is estimated that 50 Pg C or more has been lost from Abbreviations: AIC, Akaike Information Criterion; CEC, cation exchange capacity; POM, particulate organic matter; SOC, soil organic carbon; SSA, specific surface area Can. J. Soil Sci. (2014) 94: doi: /cjss

2 282 CANADIAN JOURNAL OF SOIL SCIENCE agricultural soils and therefore are required to restore soils to their prior C storage levels (Paustian et al. 1998; Lal 2004a, b). In agricultural ecosystems, land use management practices such as fertilizer application, crop rotation, plant residue return, and compost and manure inputs can be used to increase soil organic C (SOC) stocks and thus enhance soil quality (Zhang et al. 2010; Li et al. 2012; Miles and Brown 2011; Powlson et al. 2012). Most current models of SOC dynamics show linear increases in SOC stocks with increasing organic C inputs. However, it has been found that SOC storage efficiency (i.e., DSOC/Dorganic C inputs) can decrease and even approach zero with increasing organic C inputs in some agricultural sites (Campbell et al. 1991; Huggins et al. 1998; Gulde et al. 2008), which suggests the existence of an upper limit to SOC storage. These observations became the basis for the soil C saturation theory (Six et al. 2002; Stewart et al. 2007; West and Six 2007). According to the soil C saturation theory, a soil has a maximal SOC storage potential (i.e., C saturation level) and SOC storage efficiency decreases as the soil approaches its C saturation level. The determination of soil C saturation level can help identify soils with large C storage potentials (e.g., Angers et al. 2011) or the effect of topography and management on such potentials (e.g., Chan et al. 2008). The quantification of SOC storage efficiency in the process of soils approaching C saturation can be used to estimate rates and durations needed to reach maximal C storage under specific management soilclimate conditions. Organic C stabilized by fine soil particles accounts for a large proportion of the entire SOC stock (Christensen 1996; Kahle et al. 2002a), which makes the understanding of C saturation of soil minerals more meaningful. In studies in which soil C saturation has been observed, fine soil fractions were more likely to reach C saturation than coarse soil fractions or bulk soil (Hassink 1997; Gulde et al. 2008; Stewart et al. 2008, 2012). This is likely because organic matter associated with fine soil fractions is largely protected by sorption to soil minerals, and these soil minerals have finite surface areas. Thus, the amounts of organic C that can be stabilized on soil minerals may be limited by the finite mineral surface area (Hassink 1997). The amount of organic C stabilization by soil minerals can be indicated by the organic C loading, which is the amount of organic C expressed on a per unit of mineral surface area basis (e.g., mg C m 2 ). The maximal organic C loading has been proposed as 1 mg C m 2 for soils and sediments by Mayer et al. (1994). Few studies have found organic C loadings greater than 1 mg C m 2 for soils, except for organic soil horizons or light fraction samples of soils, which mainly consist of uncomplexed organic substances (Mayer and Xing 2001; Wagai et al. 2009). However, Feng et al. (2011) demonstrated that the so-called monolayer equivalent value may underestimate maximal C stabilization capacity. The amount and composition of organic matter stabilized on soil minerals is affected by soil properties associated with mineralogy (e.g., mineral specific surface area and charge characteristics) (Leinweber et al. 1993; Guibert et al. 1999; Eusterhues et al. 2005; Mikutta et al. 2005; Zinn et al. 2007). While many studies report how fertilizer or manure amendments improve soil fertility and crop productivity, and increase soil C stocks, few studies have addressed how long-term organic amendments might be used to study the factors and mechanisms regulating soil C saturation behavior. The goal of this study was primarily to test whether soil C saturation can be observed in a wide range of agricultural soils receiving long-term organic amendments, and to find which soil mineralogical properties might affect soil C saturation. Soils from seven long-term agricultural experiments were selected for this study. These soils differ substantially in edaphic factors, and represent typical soil types in the northern, middle, and southern regions in China and dominant prairie-derived cropland soils in Canada. Each experiment consists of treatments that were presumed to induce differing amounts of soil C inputs, including multiple levels of manure amendment rates. The multiple levels of soil C inputs in each experiment enable the potential observation of soil C saturation behavior, and the relatively long durations (1032 yr) of manure amendments were presumed to result in substantial C accumulations under steady state conditions. MATERIALS AND METHODS Study Sites and Sampling The primary requirement for testing soil C saturation behavior is a set of soils with differing rates of long-term organic C inputs, with all other climatic and edaphic conditions held constant. We selected several long-term agroecosystem experiments performed at six sites receiving different forms and rates of organic C additions. The experiments and sites provided a range of site variables (e.g., mean annual temperature, mean annual precipitation, soil ph, texture, and mineralogy), differing types and amounts of organic C inputs from manure and fertilizer applications, and crop rotations (Table 1). Gongzhuling, Zhengzhou and Qiyang The Gongzhuling, Zhengzhou, and Qiyang experimental sites are part of the National Soil Fertility and Fertilizer Effects Long-term Monitoring Network in China. Soils from these three sites were selected because they share the same experimental design, and represent typical soil types in the northern, central, and southern regions of China. Experimental plot sizes are 400 m 2 at Gongzhuling and Zhengzhou, and 200 m 2 at Qiyang (Xu et al. 2006). Although there is no field replication of experimental plots in these three sites, soils were randomly sampled from three different locations within each plot and kept separate for further analyses. The plots were sufficiently large to expect substantial within plot variability,

3 FENG ET AL. * C IN SOILS RECEIVING LONG-TERM MANURE 283 Table 1. Geographic location, mean annual temperature (MAT), mean annual precipitation (MAT), soil properties, fertilizer application, and crop rotation in seven agricultural experiments with long-term manure amendments. Data from Zhang et al. (2009), Wen et al. (2003) and Paré et al. (1999) Soil texture Site Location MAT (8C) MAP (mm) Soil ph Soil taxonomy Silt (g kg 1 ) Clay (g kg 1 ) Crop Gongzhuling, China N, E Typic Argiustolls Corn Zhengzhou, China N, E Typic Eutropept Wheatcorn Qiyang, China N, E Typic Hapludult Wheatcorn Laiyang, China N, E Typic Ustochrept Wheatcorn Dixon, Canada N, E Udic Boroll Oilseedcereal Melfort, Canada N, E Mollic Cryoboralf Oilseedcereal and therefore these samples were retained as pseudoreplicates rather than composited. Surface (020 cm) soil samples were collected from Gongzhuling in October 2010, from Zhengzhou in May 2011, and from Qiyang in October Prior to the initiation of the field experiments, the fields had been under continuous corn (Gongzhuling) or cornwheat (Zhengzhou and Qiyang) cropping for at least 10 yr. The fertilizer experiments were initiated in 1990, when cropping practices were continued as before, and 11 fertilizer treatments were initiated at all three sites. Soil C concentrations prior to the initiation of the experiments were 1.55% in Gongzhuling, 0.78% in Zhengzhou and 0.78% in Qiyang (Xu et al. 2006). Of the existing field treatments, the following four treatments were selected for the current study: non-fertilized as the control (CK), mineral fertilizer additions (NPK), mineral fertilizer and livestock manure additions (NPKM), and 1.5 times both mineral fertilizer and manure inputs as those in the NPKM treatment (1.5NPKM). Mineral fertilizers were added as urea or diammonium phosphate for N, as calcium superphosphate for P, and as potassium chloride or potassium sulfate for K. Fertilizer addition rates varied from site to site, but were consistent through all years of the experiment (Table 2). Manure addition rates also varied from site to site, but were designed to ensure a fixed ratio of organic to inorganic N of 7:3 at all three sites. The overall C concentration of the manure applied over the duration of the field experiment was g C kg 1 dry matter Table 2. Annual mineral fertilizer and manure application rates in seven long-term agricultural experiments Mineral fertilizers and C:N ratio of 20 (Table 2). In Gongzhuling, manure and mineral fertilizers were applied before seeding (i.e., 1/3 N, and all P and K) and the remaining 2/3 mineral N fertilizer was applied at the stem elongation stage. In Zhengzhou and Qiyang, manure and mineral fertilizers were applied before seeding (i.e., 1/2 N, and all P and K) and the remaining N mineral fertilizer was applied at wheat elongation and flowering and corn tasseling. At the Zhengzhou and Qiyang sites, annual amounts of fertilizer applied for corn growth were 3/7 of those applied for wheat growth. Laiyang The Laiyang site is located in Shandong, China, and is the first experimental site in China designed to monitor the long-term effects on crop yields of mineral and organic fertilizers applied at different rates and combinations. It has been in operation since 1978 (Yao et al. 1991). Soil C concentrations prior to the initiation of the experiment were % (Yao et al. 1991). In Laiyang, the treatments selected were the control (CK), normal manure inputs (1 ), and double manure inputs (2). The experimental plot sizes are 33.3 m 2 for each treatment, and each treatment has three field replicates. The pig manure addition rate is 8.8 Mg (dry weight) ha 1 in the (1) treatment, and the C concentration of the manure is g C kg 1 dry matter. Soil samples were collected from the top 020 cm horizon in May Manure Experiment N (kg ha 1 ) P 2 O 5 (kg ha 1 ) K 2 O (kg ha 1 ) Application rate z (Mg dry matter ha 1 ) Manure C concentration (g kg 1 ) Gongzhuling Zhengzhou Qiyang Laiyang Melfort y Dixon cattle Dixon swine y z Annual manure application rate is for the treatment with the lowest level at each experiment. y Liquid manure application rate is in units of m 3 wet matter ha 1.

4 284 CANADIAN JOURNAL OF SOIL SCIENCE Dixon and Melfort Samples were collected from two long-term manure management research sites in Saskatchewan, Canada. At each site, liquid swine manure has been applied at different rates and sequences for more than 10 yr. A similar experiment with solid cattle manure was present at the Dixon site only. The original objective of these experiments was to assess how different rates and sequences of application of liquid swine manure and solid cattle manure influence soil nutrient amounts and distribution, crop nutrition and yield (Stumborg et al. 2007). The Dixon site is located in east-central Saskatchewan, while the Melfort site is located in northcentral Saskatchewan. Both sites are on loamy Mollisolic soils. Prior to the initiation of manure inputs experiment (fall of 1996 at Dixon and fall of 1999 at Melfort), no manure had been applied. Prior to the initiation of the experiment, SOC concentrations (0- to 15-cm depth) were 2.3% at the Dixon swine manure trial, 3.3% at the Dixon cattle manure trial, and 3.5% at the Melfort trial. The experimental design at both sites is a randomized complete block design with four field replicates for each treatment, consisting of 30-m 6-m plots. At the Dixon site, liquid swine and solid cattle manure have been applied at different rates and sequences: no fertilizer as the control (CK), one time agronomic rate manure inputs (1), two times manure inputs (2), and four times manure inputs (4) (Table 2). Liquid swine manure effluent from a nearby earthen storage pit was injected 1013 cm deep using a low-disturbance injector each year. The cattle manure was broadcast and incorporated using a rotary tiller every year. Surface soil samples (020 cm) were collected in October At the Melfort site, only liquid swine manure has been applied at different rates. Surface soil samples were collected in October 2010 from three manure input rate treatments: no fertilizer as the control (CK), a one-time agronomic manure application rate each year (1 ), and three times this manure application rate made every third year (3). Liquid swine manure was applied to soils at the Melfort site in the same way as the Dixon site. Soil Fractionation Soil samples were air-dried and passed through a 2-mm sieve prior to being shipped to the University of Pennsylvania. Samples were further separated using a combined size and density fractionation scheme. Size fractionation was used to obtain the coarse soil fraction ( 53 mm) consisting of particulate organic matter (POM) and sand as a relatively labile, non-saturatable soil C pool. The fine soil fraction (B53 mm) consisted of silt and clay as a relatively stable, saturatable soil C pool. Subsequent density fractionation was used to further separate the fine fraction into the uncomplexed pool consisting of free, organic matter in particulate form in the light fraction, and the bound, organomineral pool in the heavy fraction (sensu Chenu and Plante 2006). In the size fractionation, 30 g of air-dried soil sample were combined with 10 glass beads (6 mm diameter) and 100 ml deionized water in a 500 ml Nalgene bottle, and shaken for 18 h on a reciprocal shaker (120 rpm). The suspension was poured onto a 53-mm sieve and washed until the solution passing the sieve was clear. The fraction retained on the sieve (i.e., POMsand) was washed into a pre-weighed aluminum pan and oven dried at 508C. The fraction passing through the sieve (i.e., siltclay) was freeze-dried. In the subsequent density fractionation, 6 g of freeze-dried siltclay fractions was vigorously mixed with 30 ml sodium polytungstate (density1.8 g cm 3 ) for 10 s on a vortex shaker and allowed to stand for 48 h. The light fraction and sodium polytungstate solution were siphoned, vacuum filtered and washed on a Whatman no. 42 filter, while the heavy fraction was washed with deionized water three times by centrifugation, then oven-dried. Chemical Analyses Total C and N concentrations in the bulk soil, POM sand, and siltclay were analyzed by dry combustion using a Carlo-Erba NA Samples were ground to pass through a 500-mm sieve prior to CN analysis. Soils from Gongzhuling and Zhengzhou were found to contain carbonate, and therefore organic C concentrations in these samples were determined after removing carbonates using acid fumigation (Harris et al. 2001). Samples were placed in silver capsules, moistened, fumigated with concentrated HCl for 48 h, oven-dried at 508C, and then analyzed for C concentrations. As the relatively stable and saturatable soil C pool, and therefore the fraction of primary interest, siltclay fractions were subjected to several additional analyses to generate information about the nature of the organomineral complexes and potential mechanisms for C saturation. Mineral specific surface area (SSA) of silt clay fractions was determined by the N 2 adsorption analysis. Before SSA analysis, organic matter in silt clay fraction samples was removed by placing samples in a muffle furnace in a normal atmosphere at 3508C for 18 h (Keil et al. 1997). In addition, all samples were degassed at 3258C for 4 h by N 2 and He to remove adsorbed water. Nitrogen was dosed on the surfaces at 77 K with different gas pressures in a series of batch adsorption experiments in a TriStar 3000 surface area and porosity analyzer (Micromeritics, Norcross, GA). The multi-point Brunauer-Emmett-Teller (BET) method was used to calculate SSA values under the relative pressure between 0.05 and 0.3 (Brunauer et al. 1938). The organic C loading (mg C m 2 ) was calculated by dividing C concentration (mg C g 1 ) by SSA (m 2 g 1 ), and was used as an index of the degree of C saturation (sensu Feng et al. 2011). Cation exchange capacity of the siltclay fraction was determined by the silver-thiourea method after the

5 FENG ET AL. * C IN SOILS RECEIVING LONG-TERM MANURE 285 removal of organic matter by placing samples in a muffle furnace at 3508C for 12 h (Oorts et al. 2003; Dohrmann 2006). Half a gram of freeze-dried siltclay fraction were extracted by ethanol-glycol with a 50:1 solution-to-soil ratio on a reciprocal shaker for 1 h. The suspensions were filtered (Whatman no. 42), washed with 15 ml ethanol-glycol, and air-dried on the filter papers. After air-drying, soils and filter papers were placed together in a 100-mL polyethylene bottle with 50 ml of 0.01 mol L 1 silver-thiourea, and shaken for 4 h. The extract was then filtered (Whatman no. 42) and the filtrate was stored in glass containers at 48C. The silver concentration of the filtrate was analyzed by inductively coupled plasma spectroscopy (Spectro Genesis, Mahwah, NJ). The amounts of silver sorbed to soils, indicated by the difference of silver concentrations before and after extraction, were used to calculate cation exchange capacity. Total active Fe (Fe d ) and Al (Al d ) concentrations in siltclay fractions were analyzed using the dithionitecitrate extraction method. Briefly, 0.25 g of sample was mixed with 0.5 g sodium dithionite and 25 ml sodium citrate (22%) in 50-mL centrifuge tubes and shaken for 16 h. Five milliliters of 0.05 mol L 1 MgSO 4 was added as flocculent to each sample. After centrifugation at 1520 g for 15 min, the supernatant was appropriately diluted to determine Fe/Al concentrations using inductively coupled plasma spectroscopy (Spectro Genesis, Mahwah, NJ) (Kahle et al. 2002b). Amorphous active Fe (Fe o ) and Al (Al o ) concentrations were determined using acid ammonium oxalate extraction. Briefly, 0.25 g of siltclay fraction was mixed with 25 ml ammonium oxalate (0.2 mol L 1 ) and oxalic acid (0.2 mol L 1 )at ph 3, and shaken for 4 h in the dark. The extract was centrifuged at 1520g for 30 min, and Fe/Al concentrations of the supernatant were measured by inductively coupled plasma spectroscopy (Spectro Genesis, Mahwah, NJ) (Blakemore et al. 1987). Estimation of Organic C Inputs Gross organic C inputs were estimated as the sum of manure C inputs and plant C inputs in the seven experiments. Manure C inputs were calculated using the frequency of manure application, annual manure application rates, and average C concentrations of manure applied in these experiments (Table 2). Plant C inputs were estimated by multiplying the sum of above-ground and below-ground biomass residues with C concentrations of plant residues at each experiment. Annual above-ground residues were calculated from standard regression equations to convert measured grain yields to estimate above-ground residues, and below-ground residues were estimated using ratios of below-ground residues to above-ground biomass (Intergovernmental Panel on Climate Change 2006) (Table 3). Grain yields for each crop in these seven experiments are from published technical reports or personal communications, and are relatively constant during the experimental period. At the Gongzhuling, Zhengzhou, Qiyang, and Laiyang sites in China, all above-ground plant material was removed from the field with the exception of short stubble (310 cm), and therefore plant C inputs consisted of below-ground plant C inputs only. In Gongzhuling, Zhengzhou, Qiyang, and Laiyang sites, published values (National Center for Agricultural Technology Service 1994) were used for plant C concentrations of corn (444 g Ckg 1 ) and of wheat (399 g C kg 1 ). While at the Melfort and Dixon sites, a published value (Kong et al. 2005) for plant C concentration of 430 g C kg 1 was used for plant residue of barley, oat, wheat, canola, and flax. Because C saturation behavior is observed as a function of C inputs and the saturation of the silt clay fraction is of particular interest, the net organic C inputs into the siltclay fraction on a normalized scale should be used to assess C saturation behavior. However, it is difficult to know the exact amounts of organic C inputs that are transferred into and out of the silt clay fraction. Stewart et al. (2008) demonstrated that bulk soil C concentrations can be used as a proxy for net organic C inputs, because organic C concentrations of bulk soil are the result of manure and plant residue addition to soils and decomposition of these organic C inputs. We thus used both gross C inputs and net C inputs (as bulk SOC) to test for soil C saturation behavior in response to long-term treatments. Statistical Analysis Treatment differences in the organic C concentrations of bulk soil, POMsand, and siltclay fraction were Table 3. Equations and parameters used to estimate plant C inputs to soils in seven long-term agricultural experiments in China and Canada. The regression equations estimate above-ground residue dry matter (y, Mgha 1 ) as a function of grain yield (x, Mgha 1 ) Experiment Crop Regression equation Ratio of below-ground to above-ground biomass Gongzhuling Corn y1.03x Zhengzhou Qiyang Wheat y1.61x Laiyang Melfort Barley y0.87x Dixon cattle Oat y0.81x Dixon swine Wheat y1.34x Canola and flax y1.03x

6 286 CANADIAN JOURNAL OF SOIL SCIENCE determined using ANOVA and multiple comparisons using the Tukey HSD method in JMP7.0 (SAS Institute, Inc., Cary, NC). Soil C saturation is characterized by an asymptotic relationship between soil C concentration and C inputs at steady state (Six et al. 2002; Stewart et al. 2007, 2008). In the seven agroecosystem experiments used in this study, the fertilization, residue management, and manure application treatments have been in place for more than 10 yr, and soil C concentrations were therefore considered to have approached steady state. Changes in organic C concentrations of the siltclay fraction with increasing rates of C inputs were fit to linear and asymptotic regressions. The linear regression suggests that soil fraction has not reached C saturation, and the asymptotic regression indicates that soil fraction is likely approaching or has reached C saturation. Carbon inputs used in the regression fitting were gross and net organic C inputs. The linear and asymptotic regression fitting was done by using CurveExpert Pro v1.5 (www. curveexpert.net), and slopes from the linear regression and C saturation levels from the asymptotic regression as well as Akaike Information Criterion (AIC) values for both the linear and asymptotic regressions were reported for each experiment. A model with smaller AIC value is considered to be a better fit. Linear regression of organic C concentrations with mineralogical properties [i.e., mineral SSA, cation exchange capacity (CEC), Fe d,fe o, and Al d ] for the siltclay fraction were performed in JMP 7.0 (SAS Institute, Inc., Cary, NC) to determine the influence of soil mineralogical properties on amounts of organic C associated with fine soil particles. RESULTS Bulk Soil C Responses to Long-term C Additions The various long-term agronomic treatments in each of the experiments resulted in a range of gross C inputs. Unamended treatments ranged from 1.6 Mg C ha 1 in Melfort to 27.9 Mg C ha 1 in Laiyang, while amended treatments ranged from 5.9 Mg C ha 1 in Melfort to Mg C ha 1 in Laiyang (Table 4). Gross C inputs showed a small increase in the mineral fertilizer treatments and a large increase in the treatments with manure additions when compared with the control (Table 4). The resultant organic C concentrations of bulk soil differed among experiments (with individual replicate values ranging from 5.1 to 45.8 g C kg 1 soil) and among treatments within each experiment (Table 4). Except for samples from the Zhengzhou and Melfort and Dixon swine experiments, organic C concentrations of bulk soil were significantly larger in treatments with manure amendments than the control and the treatment with mineral fertilizer application alone (Table 4). It should also be noted that soil C concentrations in the control treatments were generally smaller or similar to those at the initiation of the field experiments. Table 4. Gross organic C inputs (Mg C ha 1 ), organic C concentrations of bulk soil, POMsand, and siltclay (g C kg 1 fraction) in treatments of different fertilizer applications in seven long-term agricultural experiments. Data are mean9se Experiment Treatment Gross C inputs C concentration (bulk soil) C concentration (POMsand) C concentration (siltclay) Gongzhuling Check a a a NPK a a a NPKM b b b 1.5NPKM b b b Zhengzhou Check a a a NPK a a a NPKM a b ab 1.5NPKM a c b Qiyang Check a a a NPK a a a NPKM b b b 1.5NPKM c b c Laiyang Check a a a b b b c b c Melfort Check a a a a a a a a a Dixon cattle Check a a a b ab b b ab b b b b Dixon swine Check a a a a a a a a a a a b ac For each experiment, values in a column followed by different letters are significantly different at PB0.05.

7 FENG ET AL. * C IN SOILS RECEIVING LONG-TERM MANURE 287 Distributions of Mass and C in Physically Isolated Soil Fractions Soil mass recovery after the size fractionation to separate soils into POMsand and siltclay fractions ranged from 95 to 99% for all treatmentsite combinations (Table 5). Mass proportions of POMsand fraction differed from site to site ( %), and generally showed increases with increasing C inputs in all the experiments except a decrease for samples from Laiyang and no change for samples from Melfort (Table 5). Sample mass recovery after the density fractionation to separate siltclay into light and heavy fractions ranged from 95.0 to 101.9% (Table 5). Values above 100% suggest that the removal of polytungstate was incomplete in some samples. Density fractionation was not completed for the Dixon cattle, Dixon swine and Melfort samples because the mass proportion of the light fraction was found to account for an average of only 0.6% in the siltclay fraction (Table 5). These results suggest that even in the high manure addition rate treatments, where a large amount of uncomplexed organic matter might be expected, the amount of uncomplexed organic matter in the siltclay fraction was negligible and organic matter in the siltclay fraction can be considered to be bound to soil minerals. With the exception of the 1.5NPKM treatment from Zhengzhou, carbon recoveries after the size fractionation were significantly less than 100% (Table 5), indicating that an average of 5.9% of the initial soil C was solubilized during the fractionation. Carbon concentrations of POMsand fraction and siltclay fraction differed among sites and treatments within each of the seven experiments (Table 4), with the exception of the Melfort and Dixon swine experiments. Significantly larger C concentrations of POMsand fraction and of siltclay fraction were observed in treatments with manure amendments relative to controls and mineral fertilizer application alone (Table 4). Organic C contents of POMsand accounted for 9 39% of total SOC, and generally increased with mineral fertilizer and manure treatments (Table 4). Proportions of organic C contents of siltclay fractions differed from site to site, with the lowest proportion in Laiyang (64.5%). Moreover, organic C contents of siltclay fractions were significantly higher in treatments with manure amendments than in treatments without manure amendments only in Gongzhuling, Qiyang, Laiyang and Dixon cattle experiments (Table 4). Table 5. Mass and C recovery of the size fractionation and the density fractionation for soil samples in treatments of different fertilizer applications in seven long-term agricultural experiments. Data are mean9se Experiment Treatment Mass recovery (% of initial) Size fractionation POM mass proportion (%) C recovery (% of initial) Mass recovery (%) Density fractionation Light fraction mass proportion (%) Gongzhuling Check b NPK b NPKM a NPKM a Zhengzhou Check a NPK b NPKM b NPKM ab Qiyang Check c NPK c NPKM b NPKM a Laiyang Check a b b Melfort Check a ND z ND a ND ND a ND ND Dixon Check NA a ND ND cattle a ND ND a ND ND 4 NA z a ND ND Dixon Check a ND ND swine a ND ND a ND ND a ND ND z ND, not determined; NA: data are not available. ac For each experiment, values in a column followed by different letters are significantly different at PB0.05. Columns without letters had no significant differences.

8 288 CANADIAN JOURNAL OF SOIL SCIENCE Soil C Saturation Behavior Changes in organic C concentrations of siltclay fractions as a function of gross C inputs fit both the linear and the asymptotic regressions, except for data from Melfort and Dixon swine (Figs. 13; Table 6). The linear regression was a better fit than the asymptotic regression for changes in organic C concentration of siltclay fractions as a function of organic C inputs in the five experiments, because AIC values were lower for the linear regression than for the asymptotic regression (Table 6). Based on AIC values, changes in organic C concentration of siltclay fractions as a function of net C inputs also fit the linear regression better than the asymptotic regression for samples from all experiments except Gongzhuling and Dixon cattle (Table 6). Values of AIC were only slightly higher for the linear regression than for the asymptotic regression in Gongzhuling and Dixon cattle experiments. It is therefore inconclusive whether asymptotic regression is a better fit for samples from these two experiments. Organic C loading values for siltclay fractions greater than 1 mg C m 2 were mainly found for samples from Zhengzhou, Melfort, and Dixon cattle experiments and in treatments with high levels of organic C inputs in Laiyang and Dixon swine experiments (Fig. 4; Table 7). The highest value observed was 1.4 mg C m 2, which is higher than the proposed maximum, but well below the highest values previously reported in the literature (Wagai et al. 2009). Relationships of soil C with Mineral Properties In all seven experiments, mineral SSA did not vary greatly in different treatments within each experiment but differed between experiments and soils (Table 7). Organic C concentrations of siltclay fractions did not show linear relationships with mineral SSA except in Zhengzhou, Laiyng, and Dixon cattle (Table 8). Cation exchange capacity was much lower in Qiyang than in the other experiments, and was significantly higher in treatments with manure inputs than in treatments without manure inputs in Qiyang, Laiyang, Fig. 1. Organic C concentrations of siltclay fractions as a function of gross organic C inputs (sum of plant C and manure C inputs) and net organic C inputs (bulk SOC) in Gonngzhuling, Zhengzhou and Qiyang with long-term manure amendments.

9 FENG ET AL. * C IN SOILS RECEIVING LONG-TERM MANURE 289 Fig. 2. Organic C concentrations of siltclay fractions as a function of gross organic C inputs (sum of plant C and manure C inputs) and net organic C inputs (bulk SOC) in Laiyang with long-term manure amendments. and Dixon swine (Table 7), even though samples were treated to remove organic matter prior to analysis. Organic C concentrations of siltclay fractions showed linear regressions with CEC in Qiyang and Laiyang (Table 8). Soil Fe d and Fe o oxides were much higher in Qiyang (36.2 g kg 1 ) and Laiyang (11.6 g kg 1 ) than in the other experiments ( g kg 1 ). Soil Fe o accounted for % of Fe d in the seven experiments (Table 7). Soil Al d were high in Qiyang (12.1 mg g 1 ) and Gongzhuling (6.2 mg g 1 ), and low in the other five experiments (Table 7). Organic C concentrations of siltclay fractions showed negative linear regressions with Fe d for samples from Gongzhuling, Laiyang, Dixon cattle, and Dixon swine experiments, and generally no regressions with Fe o or Al d in the experiments (Table 8). DISCUSSION Greater C concentrations in the POMsand fraction compared with the siltclay fractions demonstrated that C inputs from manure and plant residues were more likely to accumulate in coarse size and light density fractions rather than in the mineral-bound, fine size and heavy density fractions. Conversely, in the treatments with a high manure addition rate, where a large amount of uncomplexed organic matter might be expected, increased organic C in siltclay fractions was bound to soil minerals rather than accumulated as fine-sized but uncomplexed organic matter. The lack of substantial accumulations of uncomplexed organic matter in silt clay fractions suggests that the mineral associated pool has not reached the C saturation potential. This assertion was supported by observations that siltclay fractions from the seven agricultural experiments did not exhibit an asymptotic relationship with C inputs. The results indicate that siltclay fractions still have a substantial C saturation deficit, could stabilize additional amounts of C inputs, and thus continue to act as sinks for atmospheric C. The reasons for such large soil C saturation deficits are a combination of a large C saturation potential and relatively low C inputs. Mineral SSA of siltclay fractions were relatively high in the seven agricultural experiments chosen for this study, ranging from 13.0 to 42.3 m 2 g 1 with the average of 28.9 m 2 g 1 (Table 7). The low SSA values at Zhengzhou are associated with relatively coarse soil texture (130 g clay kg 1 soil) compared with the other soils (mean of 410 g clay kg 1 soil). The fine soil texture combined with high- SSA minerals generated a large C saturation potential. In addition, the C inputs into soils in these agricultural experiments were relatively low and therefore insufficient to reduce the soil C saturation deficit. Annual C inputs in the seven agricultural experiments were generally less than 6.0 Mg C ha 1 yr 1, except in Qiyang (up to 8.5 Mg C ha 1 yr 1 ). Stewart et al. (2007) found that annual C inputs greater than 6.0 Mg C ha 1 yr 1 were needed for soils to exhibit C saturation behaviors. Moreover, while 1032 yr is long for a sustained agroecosystem experiment, the durations of manure amendments may have been too short to accumulate sufficient organic C inputs. The durations of manure amendments were only 10 yr in Melfort and 13 yr in the Dixon cattle and Dixon swine experiments. The challenge in an agronomic setting is differentiating between reaching an equilibrium SOC level after the implementation of manure additions, which West and Six (2007) estimate to be at least 21 yr, versus reaching soil C saturation, which can only be reached when higher rates of organic amendments do not yield further increases in SOC levels (Stewart et al. 2007). Estimating the required rates of C inputs to achieve soil C saturation is particularly challenging because of the lack of means to determine the maximum C stabilization capacity. Current estimates are based on comparisons of C contents of siltclay fractions with the C storage capacity determined by percentages of siltclay fractions (e.g., Hassink 1997; Angers et al. 2011), or on asymptotic regressions between C concentrations of siltclay fractions with C inputs (e.g., Stewart et al. 2008), or on evidence for a so-called

10 290 CANADIAN JOURNAL OF SOIL SCIENCE Fig. 3. Organic C concentrations of siltclay fractions as a function of gross organic C inputs (sum of plant-c and manure-c inputs) and net organic C inputs (bulk SOC) in Melfort, Dixon cattle, and Dixon swine experiments with long-term manure amendments. monolayer equivalent organic C loading (e.g., Mayer 1994). However, a recent review demonstrated important shortcomings in most of these methods (Feng et al. 2011), particularly the use of conventional linear regression to estimate the upper limit. When comparing C loadings of siltclay fractions in treatments with different levels of organic C inputs in the seven experiments, C loadings greater than 1 mg C m 2 in the Melfort, Dixon cattle, Zhengzhou, and Laiyang experiments accounted for 44% of all siltclay samples (Table 7). This suggests that the maximal C loading for soils may be greater than 1 mg C m 2. Organic matter in siltclay fractions was almost exclusively bound to soil minerals because the light fraction of siltclay generated from the density fractionation was negligible (Table 5). Contributions of uncomplexed organic matter to the C loadings of siltclay are therefore also negligible. This is an important distinction because many of the large C loading values previously reported in the literature (i.e., in the range of2 mgcm 2 ) are attributable to samples that likely have a substantial contribution of uncomplexed organic matter. In spite of this artifact in many observations, the combination of mineral-associated siltclay C loadings greater than 1mgCm 2, and linear rather than asymptotic regressions with C inputs suggest that the potential maximal C loading for siltclay fractions equivalent to 1 mg Cm 2 is inappropriate for use as a criterion to conclude the arrival of soil C saturation level without other evidence. We expected to find relationships between C concentrations of siltclay fractions and soil mineralogical properties (i.e., mineral SSA, CEC, and Fe/Al concentrations). However, no obvious relationships were generally observed in all seven agricultural experiments (Table 8). The absence of relationships between mineral

11 FENG ET AL. * C IN SOILS RECEIVING LONG-TERM MANURE 291 Table 6. Linear and asymptotic regressions of C concentrations of the siltclay fraction as a function of gross organic C inputs as the sum of plant C and manure C inputs and net organic C inputs in terms of C concentrations of bulk soil for soils in seven agricultural experiments with long-term manure amendments Linear Asymptotic Organic C inputs Experiment Slope r AICC C sat r AICC Gross C inputs Gongzhuling Zhengzhou Qiyang Laiyang Melfort NS z NS 16.3 NS NS 16.3 Dixon cattle Dixon swine NS NS 16.1 Net C inputs Gongzhuling Zhengzhou Qiyang Laiyang Melfort Dixon cattle Dixon swine z NS means statistical insignificance. SSA and C concentrations of siltclay fractions was probably because mineral SSA acts a limiting factor to influence C concentrations of siltclay fractions only when C concentrations are approaching C saturation level, but siltclay fractions in these experiments did not show C saturation behaviors (Table 6). Indeed, positive regression between SSA and C concentrations of soil fractions have been observed for sandy soils (Kiem et al. 2002), where these soil fractions can easily reach C saturation level due to small C saturation deficit. It is not surprising to find no relationships between CEC and C concentrations of siltclay fractions in these experiments, considering that organic matter in siltclay fractions that contributes to CEC was removed prior to analysis (Leinweber et al. 1993; Guibert et al. 1999). Negative or no relationships between Fe/Al hydroxide concentrations and C concentrations of siltclay fractions in these experiments might be caused by high affinity of the organic matter for crystalline mineral surfaces (Christensen 1992; Amelung et al. 1998; Schulten and Leinweber 2000; Zinn et al. 2007). C concentrations of siltclay fractions apparently fell below the C saturation limit, and were therefore not limited by mineral SSA. Fe/Al hydroxides are frequently large contributors to mineral SSA and therefore no, or negative, relationships with organic C concentrations are consistent with expectations. Fig. 4. Carbon loadings of siltclay fractions as a function of gross organic C inputs in seven agricultural experiments with long-term manure amendments. CONCLUSION In the seven agricultural experiments with long-term multiple levels of manure amendments, C concentrations in bulk soil, POMsand, and siltclay differed among treatments within each site and among experiments and soils. Manure inputs generally increased C in POMsand fractions more than in siltclay fractions.

12 292 CANADIAN JOURNAL OF SOIL SCIENCE Table 7. Mean of cation exchange capacity (CEC, cmol kg 1 ), total (Fe d,mgg 1 ) and amorphous active (Fe o,mgg 1 ) iron oxides, total aluminum active oxides (Al d,mgg 1 ), soil mineral specific surface area (SSA, m 2 g 1 ), and C loading (mg C m 2 ) of the siltclay fraction in seven agricultural experiments with long-term manure amendments Experiments Treatments CEC Fe d Fe o Al d SSA C loading Gongzhuling Check 39.7a 7.6ab 2.3a 6.6ab 37.2a 0.38b NPK 31.5a 8.1a 2.7a 7.1a 42.3a 0.32b NPKM 37.2a 6.8b 2.4a 5.6b 32.6a 0.60a 1.5NPKM 37.6a 7.0ab 2.5a 5.6b 37.3a 0.57a Zhengzhou Check 38.9a 5.1b 1.3b 3.6b 13.4a 1.18a NPK 44.3a 5.7a 1.6a 3.9a 13.0a 1.26a NPKM 44.2a 5.8a 1.8a 3.9a 13.9a 1.33a 1.5NPKM 40.3a 5.7a 1.7a 3.8ab 14.4a 1.39a Qiyang Check 15.3b 33.3a 4.9b 11.1b 34.5a 0.23c NPK 14.4b 38.1a 5.4ab 12.6ab 35.1a 0.27bc NPKM 19.0a 38.9a 5.8ab 12.9a 36.1a 0.34b 1.5NPKM 19.0a 34.4a 6.2a 12.0ab 32.1a 0.49a Laiyang Check 28.0b 12.9a 4.3a 5.4a 24.5a 0.40c ab 11.2ab 3.9a 4.8a 20.9b 0.85b a 10.6b 4.0a 4.9a 21.6b 1.21a Melfort Check 31.1a 8.0b 4.2a 4.7a 28.8a 1.15a a 8.5ab 3.7a 5.0b 30.0a 1.13a a 9.0a 3.9a 5.3b 30.7a 1.17a Dixon cattle Check 36.3a 7.6a 2.4b 5.5a 36.6a 0.87b a 6.3b 3.1a 4.4a 31.2ab 1.21a a 6.0b 2.9ab 4.3a 28.8b 1.26a a 6.8ab 2.7ab 5.0a 28.4b 1.40a Dixon swine Check 37.0a 7.0a 2.8a 5.0b 29.4a 0.95ab a 7.1a 2.4a 5.0b 32.5a 0.79b a 7.2a 2.4a 5.0ab 32.5a 0.81b a 5.9a 2.6a 4.0a 29.0a 1.29a ac For each experiment, values in a column followed by different letters are significantly different at PB0.05. Siltclay fractions in these experiments did not exhibit C saturation behavior, as demonstrated by a lack of asymptotic behavior as a function of C inputs. This was likely because current organic C inputs levels were insufficient for siltclay fractions to meet the large soil C saturation deficits of these agricultural soils. Moreover, the maximal organic C loading, presumed to be 1 mg C m 2, may not be true for all types of soils, as 44% of all siltclay fractions were found to have greater values. Overall, agricultural soils subjected to long-term organic C inputs at typical agronomic levels are probably still far from C saturation and, therefore, can continue to act as C sinks to stabilize more organic C inputs. Sandy soils with smaller C saturation deficits (which were not part of the current study) and sufficiently large C inputs may be the soil types in which soil C saturation behavior is more likely to be observed in the field, and should be targeted for further study to observe soil C saturation behavior and estimate the upper limit of soil C stabilization. Table 8. Linear regressions among organic C concentration, soil mineral specific surface area (SSA, m 2 g 1 ), mass proportions of siltclay fractions (siltclay, mass%), cation exchange capacity (CEC, cmol kg 1 ), total and amorphous active iron oxides (Fe d /Fe o,mgg 1 ), and total active aluminum oxides (Al d,mgg 1 ) of siltclay fractions in seven agricultural experiments with long-term manure amendments Variables Variables Gongzhuling Zhengzhou Qiyang Laiyang Melfort Dixon cattle Dixon swine C concentration SSA NS z 0.273() y NS 0.404() NS 0.344() NS CEC NS NS 0.513() 0.627() NS NS NS siltclay(mass%) 0.731() NS 0.915() 0.757() NS 0.272() NS Fe d 0.400() NS NS 0.549() NS 0.328() 0.284() Fe o 0.569() NS 0.603() NS NS NS NS Al d NS NS NS NS NS NS 0.287() SSA siltclay(mass%) NS NS NS 0.765() NS NS NS Fe d 0.795() NS 0.396() 0.594() 0.353() 0.191() 0.616() Fe o 0.352() NS NS 0.590() NS NS NS Al d 0.547() NS NS 0.520() NS NS NS z NS means statistical insignificance. y Values are adjusted R 2 ; / means positive/negative linear regression.