Estimating soil particulate organic carbon through total soil organic carbon

Natalia Veronica Diovisalvi1*; Guillermo Alberto Studdert2; Nahuel Ignacio Reussi Calvo1-2; Germán Franco Dominguez2; Angel Berardo1

1 Laboratorio de Suelos Fertilab;
2 Facultad de Ciencias Agrarias - UNMdP, Unidad Integrada Balcarce. *Autor de contacto: ndiovisalvi@laboratoriofertilab.com.ar

Recibido: 27-07-13
Recibido con revisiones: 08-02-14
Aceptado: 15-02-14

 

---

ABSTRACT

Particulate soil organic carbon (POC) is sensitive to management changes, and a good soil health indicator. However, its determination is tedious and time consuming. Given POC changes represent most of total organic carbon (TOC) variation due to soil use, we hypothesized that POC can be estimated through the variation of TOC. Our objective was to evaluate changes of POC and TOC for a wide range of soil silt (Si) + clay (Cy) contents to assess the feasibility of estimating POC through TOC. Soil samples (n=161) were collected from the surface layer of fields under no tillage throughout the Argentinean Pampean Region, and their TOC and POC concentrations were determined. The variation of POC was described with ex-ponential models fitted as a function of both TOC and TOC/(Si+Cy). The models fitted adequately and estimated POC as a function of TOC (R²=0,96, observed vs estimated not different from the 1:1 line). No improvement was observed when POC and TOC contents were relativized to Si+Cy contents (R²=0,97). Approximately 74% of the estimated values of POC deviated less than ±10% from the observed values. We conclude that for the conditions studied, the model acceptably estimated POC knowing TOC hence saving time and resources to assess soil health status.

Key words. Organic matter fractions, soil health indicator, soil texture, modelling.

ESTIMACIÓN DEL CARBONO ORGÁNICO PARTICULADO A TRAVÉS DEL CARBONO ORGÁNICO TOTAL DEL SUELO

RESUMEN

El carbono orgánico particulado (POC) es sensible a los cambios producidos por las prácticas de manejo, resultando un buen indicador de la salud del suelo. Sin embargo, su determinación es tediosa y consume mucho tiempo. Dado que los cambios en POC debido al uso del suelo representan la mayor parte de las variaciones del carbono orgánico total (TOC), se hipotetizó que el POC se puede estimar a través de la variación de TOC. Nuestro objetivo fue evaluar los cambios de POC y TOC en suelos con un amplio rango de contenidos de limo (Li) + arcilla (As) para evaluar la posibilidad de estimar POC a través de TOC. Se recolectaron muestras (n=161) superficiales de lotes bajo siembra directa en la Región Pampeana Argentina, y se les determinó el TOC y el POC. La variación de POC en función de TOC y TOC/(Li+As) fue ajustada con un modelo exponencial. El modelo explicó adecuadamente la variación de POC en función de TOC (R²=0,96, observado vs estimado no diferente de la línea 1:1). No se observaron mejoras cuando los contenidos de POC y TOC fueron relativizados con los contenidos de Li+As (R²=0,97). Aproximadamente el 74% de los valores estimados de POC se desviaron menos de ±10% de los valores observados. Se concluye que para las condiciones estudiadas, el modelo estimó de modo aceptable el POC conociendo el TOC, por lo tanto se ahorraría tiempo y recursos para evaluar el estado de salud del suelo.

Palabras clave. Fracciones de materia orgánica, indicador de salud de suelo, textura del suelo, modelación.

---

 

 

INTRODUCTION

Soil organic carbon (TOC) is a key soil component for agroecosystem functioning due to its role as a nutrient reservoir and supplier and in soil structure stabilization, among others. However, TOC content is dynamic and strongly affected by management practices that influence carbon (C) loss and sequestration, and dependent on the balance between C gains and losses as affected by soil management (Janzen, 2006).

The Pampean Region of Argentina (30° to 40º S and 57º to 66º O) (Fig. 1) is recognized as one of the most important grain production regions in the world (Álvarez & Steinbach, 2009). In the last decades, the progressive increase of cropping has produced a negative balance of C (Sainz Rozas et al., 2011), nitrogen (N), and other nutrients, with a concomitant deterioration of soil properties increased erosion (Manuel-Navarrete et al., 2009). Despite the steadily increase since the 90's of the cropping area under no-tillage (NT) (MAGyP, 2011), the increase of the frequency of soybean (Glycine max (L.) Merr.) in crop sequences because of its higher profitability, has counterbalanced the expected beneficial effect of NT (Manuel-Navarrete et al., 2009) due to the low amount of C returned to the soil as residues (Studdert & Echeverría, 2000). Soil health indicators are needed to develop management decisions to deal with the delicate equilibrium regulating soil functioning and with the need of a profitable and sustainable agriculture.

The cropping effects on soils can be monitored through the changes in TOC or its fraction contents (Diovisalvi et al., 2008). Particulate organic C (POC) has been proposed as an early indicator of the effects of soil management practices. It shows the size of the easily mineralizable soil organic fraction compartment, and hence is very important to estimate soil N mineralization potential (Fabrizzi et al., 2003; Gregorich et al., 2006). Likewise, POC is closely related to soil structure and stabilization, since its decomposition by microorganisms produces organic compounds that stimulate aggregation, increase macroaggregate stability, and promote long term C stabilization in microaggregates (Six et al., 2004). Therefore, the knowledge of POC content variation would allow predicting soil conditions and crop behavior related to soil management changes (Diovisalvi et al., 2008) and would be an important tool to help make management decisions to stop and revert soil health deterioration.

Figure 1. Argentinean Pampean Region map indicating isohyets and isotherms. Dots show the places from where soils samples were taken.
Figura 1. Mapa de la Región Pampeana Argentina indicando isohietas e isotermas. Los puntos indican los lugares en donde fueron tomadas las muestras de suelo.

Even though POC determination is simple, the technique is tedious and time-consuming. Therefore, its use in commercial soil laboratories as a routine method is limited and expensive for producers. It would be important to have an easily measurable estimator of POC content. Results from soils of the Argentine Pampas Region with a wide range of textures show that, in fields where erosion is not important, changes in TOC related to soil use are mainly explained by POC content variations (Diovisalvi et al., 2008; Quiroga et al., 2008; Irizar, 2010). Therefore, it would be feasible to estimate POC content using TOC content and thus avoiding fractionation.

A strong correlation between silt (Si) plus clay (Cy) (Si+Cy) content and TOC has been reported (Quiroga et al., 2006; Zinn et al., 2007), both in cropped and in undisturbed soils (Hassink, 1997; Hevia et al., 2003). Hence, TOC content and C sequestration capacity increase with finer textures (Six et al., 2002; Zinn et al., 2007). The greater the Cy content, the larger the specific surface where humified C can be adsorbed and protected (chemical protection; Six et al., 2002) and therefore, the greater the amount of C associated to the mineral fraction (AOC). Besides, higher Cy content is related to more aggregation, aggregate stability and physical protection of both AOC and POC making it less exposed to mineralization (Six et al., 2002). Hence, soil texture affects the rate and magnitude of TOC fraction changes due to soil use. In order to account for the differential protection of soil organic C associated to texture, indexes like TOC/(Si+Cy) have been proposed to analyze and compare TOC (Quiroga et al., 2006) and its fraction contents and variation among soils with different particle size distributions.

We hypothesize that since POC changes represent most of the variation of TOC, POC changes associated to soil use can be estimated through the variation of TOC. Our objective was to evaluate changes of POC and TOC throughout a wide range of soil textures in order to assess the feasibility of estimating the former knowing the latter.

MATERIALS AND METHODS

Soil samples and sample processing

This study was performed on 161 soil samples from the arable layer (0-20 cm) of fields under NT with different management histories and years under continuous cropping, ranging between null disturbance by cropping and more than 25 years of continuous cropping. Soil samples were collected from different areas of the Pampas Region in Argentina (Figure 1): Southeastern, Southwestern, Central, Northeastern and Northwestern areas of Buenos Aires Province, South-western area of Entre Ríos Province, Southern area of Santa Fe Province, and Southeastern area of Córdoba Province. The climate is temperate-humid (18 °C isotherm in the north through 14 °C isotherm in the south) (Figure 1). Its topography is flat to slightly rolling and its soils are predominantly Mollisols originating mainly from loessic sediments and to a lesser extent from fluvial sediments. Therefore, these soils show a wide range of depths, textures, organic matter contents and fertility (Álvarez & Lavado, 1998). In general, soil texture is finer from the Southwest (SW) to the Northeast (NE), coinciding with annual rainfall and mean temperature increments (Figure 1), and the particle size distribution of the parent material (coarser in the SW and finer in the NE). Hence, sandy and sandy-loam Typic and Entic Haplustolls predominate in the West and SW regions, whereas loam, silt-loam and silty-clay-loam Typic and Vertic Argiudolls predominate in the East and NE (Duran et al. , 2011).

The contents of TOC, AOC, and POC were determined on each soil sample. The mineral associated fraction (<50 μm) was separated through the fractionation method described by Cambardella & Elliott (1992). Whole soil organic C (TOC) and AOC contents were determined by wet combustion through complete oxidation with potassium dichromate in solution with sulfuric acid maintaining a reaction temperature of120 °C during 90 min (Schlichting et al. , 1995). Particulate organic C content was calculated by subtracting AOC content from TOC content (Cambardella & Elliott, 1992).

The proportion of the different size mineral fractions was determined through a modification of the method described by Gee & Bauder (1986). We used a constant soil aliquot (50 g) regardless of soil texture, instead of 10-20 and 60-100 g for fine-textured soils and coarse-textured soils, respectively, as proposed by Gee & Bauder (1986). Soil samples were grouped in four arbitrary texture classes according to their Si+Cy contents: 1) Fine (>750 g kg^-1 Si+Cy) (Fi), 2) Loam fine (>500 g kg^-1 - <750 g kg^-1 Si+Cy) (LFi), 3) Loam coarse (>250 g kg^-1 - <500 g kg^-1 Si+Cy) (LCo), and 4) Coarse (<250 g kg^-1 Si+Cy) (Co). Nine, 79, 53, and 20 soil samples out of the 161 samples corresponded to the Fi, LFi, LCo, and Co textural classes, respectively.

Model development and validation

Variations in POC and POC/(Si+Cy) were related to variations in TOC and in TOC/(Si+Cy) through exponential models. Curve fitting was performed through non- linear regression analysis (R Development Core Team, 2009). The model used to fit POC and POC/(Si+Cy) as a function of TOC or TOC/(Si+Cy) was:

DVp = -I + I * exp(r * IV)                                       (Eq. 1)

where DVp is the dependent variable (POC (g kg^-1) or POC/ (Si+Cy) (g kg^-1/ g kg^-1)), I is the intercept, r is the exponential rate of change (1/(g kg^-1) or 1/(g kg^-1/ g kg^-1)) and IV is the independent variable (TOC (g kg^-1) or TOC/(Si+Cy) (g kg^-1/ g kg^-1)).

In order to perform the curve-fitting and then the corresponding model performance evaluation, the data set (n=161) was randomly split into two sub-sets: 70% of the data was used for curve-fitting and the remaining 30% was used for model validation. When splitting the data set, each textural group was proportionally represented in the resulting sub-sets (fitting and validation). Random data splitting, curve fitting, and model validation were run five times and the model that best performed for each variable relationship was selected.

Model performance evaluation

Several statistic tools were used to evaluate the agreement between the observed and the estimated values of each dependent variable (DV) after the validation process and to choose the model that best described the variation of the DV as a function of the IV. On the one hand, linear simple least square regression analyses of DV observed values on DV estimated values were performed. The equality of the intercept and of the slope to zero and one, respectively, was tested through F-tests both separately and simultaneously.

On the other hand, we calculated the difference between DV observed values and DV estimated values to obtain the bias error (BE), and we calculated the relative error (RE, %) as the quotient between BE's and their corresponding observed values. The BE's an RE's means were referred to as the mean bias error (MBE) and mean relative error (MRE, %), respectively. Root mean squared variation (RMSV) (Kobayashi & Salam, 2000) and root mean squared error (RMSE) (Fox, 1981), were also calculated. All calculations and analyses for model performance evaluation were done on the computational environment of R (R Development Core Team, 2009).

The best models were those with the highest R² both in the curve fitting step and in the linear regression between observed and estimated DV values. Likewise, the coefficients of these

linear regressions (intercept and slope) had to be significantly equal to 0 and 1, respectively, both separately and simul-taneously. On the other hand, the best model had to have generated the lowest RMSE and RMSV. Normality (Shapiro-Wilk test) and homocedasticity (visual distribution in plot as a function of observed values) of BE (R Development Core Team, 2009) were checked in all cases.

Re-validation with independent data

Soil-surface layer (0-20 cm) TOC and POC contents determined annually between 1994 and 2006 in samples from the a long-term experiment carried out at the Unidad Integrada Balcarce, Balcarce, Buenos Aires Province (37°45'S, 58°18'W, 138 m above sea level), were used to re-validate the selected best models. The soil is a complex of fine, mixed, thermic Typic Argiudoll (USDA, 2006) (Mar del Plata series, INTA, 1979) and fine, illitic, thermic Petrocalcic Paleudoll (USDA, 2006) (Balcarce series, INTA, 1979) with less than 2% slope (low erosion). Particle size distribution of soil complex was 230, 360, and 410 g kg^-1 Cy, Si and Sa contents (INTA, 1979), respectively, and the textural class was LFi (>500 g kg^-1 - d»750 g kg^-1 Si+Cy). The experimental design was described by Cozzoli et al. (2010). Briefly, the cropping systems included: continuous cropping under NT and under conventional tillage (CT), 25% of the rotation cycle (3 yr) with pasture followed by 75% of the rotation cycle (9 yr) with cropping under NT and under CT, 50% of the rotation cycle (3 yr) with pasture and 50% of the rotation cycle (3 yr) with cropping under NT and under CT, and continuous pasture. Soil TOC and POC contents were determined as described above. Model performances, normality and homoscedasticity were evaluated with the same statistical tools mentioned before.

RESULTS

Table 1 shows the means and coefficients of variation (CV) of Si+Cy, Sa, TOC, AOC and POC contents for all four texture classes defined. It is worth pointing out that the largest number of samples considered for this study was from the medium texture classes (LFi and LCo) which in turn correspond to the most intensively cultivated area of the Argentine Pampas Region.

Curve fitting and model validation with validation data set

In all cases, the exponential model shown as Eq. 1 adequately described (P < 0,0001) the variation of the DV as a function of the IV, with very high R² (all > 0,88, data not shown). The lowest R² (between 0,89 and 0,91, data not shown) were achieved by the model fitted to POC as a function of TOC/(Si+Cy) with Eq. 1, whereas the rest of the relationships had R² above 0,95 (data not shown). However, the validation step showed performance differences among the models generated by the five curve-fitting runs for each DV vs. IV relationship (data not shown). Table 2 shows the results of curve-fitting and model performance of the best model for each DV vs. IV relationship.

Table 1. Soil variables for each textural class. Tabla 1. Variables de suelo para cada clase textural.

Texture	n(a)	Si+Cy«		Sa	(a)	TOC	(a)	AOC	a)	P0C«
class		Mean	CV(a»	Mean	CV«	Mean	CV«	Mean	CV«	Mean CV(a)
	9					g kg4
Fine		839,4	10,1	160,6	52,8	26,8	31,2	17,9	31,7	8,9 39,2
Loam Fine	79	571,0	10,5	429,0	14,0	36,9	23,6	24,2	18,6	12,7 43,1
Loam Coarse	53	387,0	17,1	613,0	10,6	26,9	50,0	16,4	43,1	10,6 70,8
Coarse	20	192,9	22,1	807,1	5,3	11,8	46,2	7,3	45,3	4,5 58,7

^(a) n: number of soil samples, Si+Cy: silt plus clay content, CV: coefficient of variation, Sa: sand content, TOC: total soil organic carbon, AOC: mineral associated soil organic carbon, POC: particulate soil organic carbon.

 

As mentioned before, the poorest performance was shown by the models describing the relationship between POC and TOC/(Si+Cy). Even though the R² achieved for this relationship was high (0,90, Table 2), the variation of the estimated values of POC did not explain the variation in the corresponding observed ones adequately (low linear regression R², Table 2 Model validation). That model tended to overestimate POC as a function of TOC/(Si+Cy) since MRE was negative indicating that, on average, estimated values were higher than the observed ones. It is worth remarking that the highest negative MRE values were obtained for the coarser texture classes being -120,7% for Co, and -36,9%, for LCo.

Table 2. Curve- fitting (Equation 1) step and model validation step results.
Tabla 2. Resultados del ajuste del modelo (Ecuación 1) y de su validación.

Independent variable
T0Cw		T0C/(Si+Cy)«
Performance statistics	Dependent variable
P0C«	P0C«	P0C/(Si+Cy)«
Curve fitting
P > F <0,0001	<0,0001	<0,0001
R2 0,96	0,90	0,97
Model validation
Simple linear regression observed	vs. estimated valúes
P > F (l=0)« 0,60	0,41	0,29
P > F (S=1)« 0,65	0,38	0,19
P > F (l=0 and S=l)(a) 0,55	0,50	0,42
R2 0,88	0,65	0,84
Statistics based on the difference between observed and estimated valúes
MBEw(g kg4) -0,18	-0,03	0,00
MRE(a) (%) -2,40	-21,87	-3,36
RMSE« 2,15	3,67	0,04
RMSV« 2,15	3,67	0,04

^(a) TOC: total soil organic carbon, AOC: mineral associated soil organic carbon, POC: particulate soil organic carbon, Si+Cy: silt plus clay content, I: intercept, S: slope, MBE: mean bias error, MRE: mean relative error, RMSE: root mean squared error, RMSV: root mean squared variation.

Table 2 shows that the models to estimate POC as a function of TOC, (Fig. 2a) and POC/(Si+Cy) as a function of TOC/(Si+Cy) (Fig. 2b) performed best. Variation of the IV (TOC or TOC/(Si+Cy)) explained more than 95% of the variation of the DV (POC and POC/(Si+Cy)). Likewise, in the validation step, the variation of the estimated values explained most of the variation of the observed ones (R² of the simple linear regressions were 0,88 and 0,84, respectively). Table 2 and Figure 3 also show that the linear regression models of observed values versus the estimated ones in the validation step were statistically equal to the 1:1 line.

The two models obtained (Fig. 2) tended to slightly overestimate the DV values since MRE values were negative (-2,40 and -3,36%, Table 2), indicating that, on average, estimated values were higher than the observed ones. However, only 10 and 28 % of the estimated values in the validation step showed RE within ±5%, and 18 and 41% within ±10%. On the other hand, the MRE by texture classes indicate that the models showed the poorest performance for both Fi and Co texture classes (-18,6% and 25,5%, and 34,0% and 23,3%, respectively). However, it is worth to recall that the number of cases of these two textural classes included in this study was low (n=9 and n=20, respectively, Table 1) and could have not been representative of a wide enough range of management situations to adequately take into account the real variability.

Figure 2. Results of fitting equation 1 to particulate soil organic carbon (POC, a) as a function of total soil organic carbon (TOC) content and to POC relative to silt + clay contents (POC/(Si+Cy), b) as a function of TOC relative to silt + clay contents (TOC/(Si+Cy)).
Figura 2. Resultados del ajuste de la ecuación 1 para los contenidos de carbono orgánico particulado (POC, a) en función de los contenidos de carbono orgánico total (TOC) y para los contenidos de POC relativizados por los contenidos de limo + arcilla (POC/(Si+Cy), b) en función de los contenidos de TOC relativizados por los contenidos de limo + arcilla (TOC/(Si+Cy)).

Figure 3. Results of simple linear regressions between observed (obs.) values in the validation data set of particulate soil organic carbon (POC, a) and POC relative to silt + clay contents (POC/ (Si+Cy), b), and the estimated (est.) values with the corresponding models shown in Figure 2. Results of statistical analysis of the regressions are shown in Table 2.
Figura 3. Resultados de la regresión lineal simple entre los valores observados (obs.) en la validación con el set de datos de carbono orgánico particulado (POC, a) y POC relativizado por el contenido de limo + arcilla (POC/(Si+Cy), b), y los valores estimados (est.) con el correspondiente modelo mostrado en la Figura 2. En la tabla 2 se muestran los resultados de la evaluación estadística de las regresiones.

 

According to these results (Fig. 2 and Table 2), for the range of textures explored, POC could be satisfactorily predicted with TOC content, especially for soils with loam texture.

Validation with an independent data set

Figure 4 shows the linear regression between the observed values of POC as a function of the estimated ones. This Figure indicates that the model obtained in the curve-fitting step (Fig. 2a) adequately estimated the changes of POC as a function of TOC. Most of the variation of the observed values was explained by the variation in the estimated ones; the R² of the linear regressions was 0,67

(Fig. 4). The model tended to overestimate POC since MBE and MRE were slightly negative (-0,33 g kg^-1 and -4,05%, respectively), but 50,5 and 74,2% of the RE for POC were within the ranges of ±5% and ±10%, respectively. This means that a great proportion of the estimated values diverged no more than 10% from the corresponding observed ones.

It is worth remarking that both conditions of slope equal to 1 and intercept equal to 0 were not achieved by the linear regressions between observed and estimated values (Fig. 4). The very narrow range of values of POC and TOC in the experiment could have been associated to this apparently not acceptable performance indicator. Values of POC and TOC employed for this validation ranged only between 6.9-13.0 g kg^-1, and 26.2-36.3 g kg^-1, respectively. These ranges are much narrower than those of the data used for model fitting and first validation (Table 2). Anyway, POC est. (POCe) (g kg-¹) the rest of the indicators (RMSE=0,83 and RMSV=0.76) denote a good performance of the model fitted to predict POC as a function of TOC (Fig. 2a).

Figure 4. Results of simple linear regressions between observed (obs.) values of particulate soil organic carbon (POC) from a crop rotation and tillage system experiment, and the estimated (est.) values with the model shown in Figure 2a. I: intercept. S: slope.
Figura 4. Resultados de la regresión lineal simple entre valores observados de carbono orgánico particulado (COP) de un ensayo de rotaciones de cultivos y sistemas de labranza, y valores estimados (est.) con el modelo mostrado en la Figura 2a. I: ordenada al origen. S: pendiente.

DISCUSSION

Representativeness of the studied soils

The total organic C content range of the soils included in this study (Table 1) is similar to that reported by Sainz Rozas et al. (2011) who analyzed TOC content changes in a wide range of soils of the Argentine Pampas Region, and by Puget & Lal (2005) for soils from several parts of the world. The CV of organic C fractions could be attributed not only to particle size distribution within each texture class, but also to the effect of different factors such as cropping history, crop sequence and fertilization which are closely related to soil organic C content. The highest TOC contents were observed in the fine to medium texture classes (Fi, LFi and LCo, Table 1). This agrees with McConkey et al. (2003) who reported higher organic C stocks under NT in fine textured soils, which could be attributed to the relationship between Cy content and to the chemical and physical protection of soil organic C (Six et al., 2002). On the other hand, cropping period duration (Studdert et al., 1997; Irízar, 2010) and crop sequence (Domínguez et al.,

2009) also affect soil organic C stock through the amount, quality and timing of C input to the soil. The larger the amount of residues returned to the soil, the higher the C input and the lower the effect of soil use on soil organic C stock (Domínguez et al., 2009; Varvel & Wilhelm, 2010). It is worth remarking that the CV of TOC and of AOC were very similar throughout all texture classes (Table 1), whereas those of POC were much higher. This confirms that POC is much more sensitive to different soil management practices and uses.

The wide range of mean TOC and its fractions contents within and among textural classes, together with the dispersion of Si+Cy and Sa contents (Table 1), indicate that soil samples used for this study represent a wide range of situations. Therefore, it is considered that the results of this studyare representative of the Argentine Pampas Region.

Prediction of POC

The results shown in Table 2 indicate that the estimation of POC as a function of TOC/(Si+Cy) would not be feasible, especially for coarse textured soils (data not shown). On the other hand, models fitted in the curve-fitting step adequately estimated POC as a function of TOC, and POC/ (Si+Cy) as a function of TOC/(Si+Cy) (Table 2).

Plante et al. (2006) observed that the relationship of texture with soil organic C differences among soils was higher for those fractions protected within microaggre-gates than for those labile or hydrolysable ones not occluded within them. This is probably due to the close relation between soil C and Si+Cy content (Hassink, 1997) which in turn also influences micoraggregate formation and stabilization (Six et al., 2002). Likewise, McLauchlan (2006) reported that soil organic C increase with the increase in Cy content was not generalized and that only the organic C fraction chemically protected by Cy (Six et al., 2002) followed that trend. Similarly, Liang et al. (2009) and Zinn et al. (2007) found that soil organic C content was strongly affected by Si+Cy content, especially Cy content. Thus, as proposed by Quiroga et al. (2006), the relativization of organic C fraction contents to Si+Cy contents to compare organic C dynamics among soils with different textures would be necessary to account for the influence of the mineral fraction on C dynamics.

As discussed before, in this study, the relativization IV (TOC/(Si+Cy)) only did not yield a good estimation of POC (Table 2). Silt+Cy content influences not only the TOC content but also TOC fraction proportions (Plante et al., 2006) and, therefore, partial relativization may have not taken such effect completely into account. However, no improvement in model performance was observed when POC and TOC contents were relativized to Si+Cy contents, when compared to the performance of the model obtained with the not-relativized variables (Figs. 2 and 3, Table 2).

Some other factors could influence the relationship between soil organic C fractions and Si+Cy content such as Cy type, climate factors such as temperature and precipitation, and management practices, among others. Hevia et al. (2003) reported that in soils from the Argentine semiarid region, POC/(Si+Cy) increased with increasing annual rainfall whereas it decreased with mean temperature increases. On the other hand, Percival et al. (2000) observed in New Zealand that neither Cy content nor climate factors explained the variation of soil organic C, while the proportion of alophans in the Cy fraction explained a great part of it. They concluded that Cy mineralogy highly influences the relationship between Cy content and organic C variation. Furthermore, Hao & Kravchenko (2007) reported that the relationship between organic C fractions and Si+Cy content under NT was much stronger than under CT. Chivenge et al. (2006) informed that in fine-textured soils, conservation practices increased soil organic C due to a reduction of its mineralization rate, while in coarse-textured ones the increase was due to a greater C input. For soils within the LFi texture class, Roldán (2011) reported that higher Cy contents were associated with greater organic C protection and macroaggregate stability regardless of the tillage system.

The soils used in this study represented a wide range of management histories, climatic conditions, and textures (Table 1, Fig. 1). The lack of improvement of model performance when variables were relativized (Fig. 2 and Table 2) indicate that the incidence of climatic and management factors on organic C differences across soils was much stronger than the Si+Cy content effect.

CONCLUSION

The results obtained in this study indicate that for the conditions studied, the fitted model satisfactorily estimated POC using TOC content, especially for soils with loam texture. This reveals a great simplification to obtain POC and with it, estimate some other soil properties important to evaluate soil quality (e.g. N mineralization capacity, aggregate stability). However, even though these results are applicable to the most cropped areas, it is necessary to deepen the research in coarser and finer textured soils. This will contribute to improve the model performance and to have a better representation of most soils, being also very important in terms of their contribution to global crop production.

REFERENCES

1. Álvarez, R & RS Lavado. 1998. Climate, organic matter and clay content relationships in the Pampa and Chaco soils, Argentina. Geoderma 83: 127-141.         [ Links ]
2. Álvarez, R & HS Steinbach. 2009. A review of the effects of tillage systems on some soil physical properties, water content, nitrate availability and crops yield in the Argentine Pampas. Soil Tillage Res. 104: 1-15.         [ Links ]
3. Cambardella, CA & ET Elliott. 1992. Particulate soil organic matter. Changes across a grassland cultivation sequence. Soil Sci. Soc. Am. J. 56: 777-783.         [ Links ]
4. Chivenge, PP; HK Murwira; KE Giller; P Mapfumo & J Six. 2006. Long-term impact of reduced tillage and residue management on soil carbon stabilization: Implications for conservation agriculture on contrasting soils. Soil Tillage Res. 94: 328-337.         [ Links ]
5. Cozzoli, MV; N Fioriti; GA Studdert; GF Domínguez & MJ Eiza. 2010. Nitrógeno liberado por incubación anaeróbica y fracciones de carbono en macro- y microagregados bajo distintos sistemas de cultivo. Ciencia del Suelo 28: 155-167.         [ Links ]
6. Diovisalvi, NV; GA Studdert; GF Domínguez & MJ Eiza. 2008. Fracciones de carbono y nitrógeno orgánicos y nitrógeno anaeróbico bajo agricultura continua con dos sistemas de labranza. Ciencia del Suelo 26:1-11.         [ Links ]
7. Domínguez, GF; NV Diovisalvi; GA Studdert & MG Monterubbianesi. 2009. Soil organic C and N fractions under continuous cropping with contrasting tillage systems on mollisols of the southeastern Pampas. Soil Tillage Res. 102: 93-100.         [ Links ]
8. Durán, A; H Morrás; G Studdert & X Liu. 2011. Distribution, properties, land use and management of Mollisols in South America. Chin. Geogra. Sci. 21: 511-530.         [ Links ]
9. Fabrizzi, KP; A Morón & FO García. 2003. Soil carbon and nitrogen organic fractions in degraded vs. non-degraded Mollisols in Argentina. Soil Sci. Soc. Am. J. 67: 1831-1841.         [ Links ]
10. Fox, DG. 1981. Judging air quality model performance: a summary of the AMS workshop on dispersion models performance. Bull. Am. Meteorol. Soc. 62: 599-609.         [ Links ]
11. Gee, GW & JW Bauder. 1986. Particle size analysis. In: GR Blake, KH Hartge (eds) Methods of Soil Analysis, Part 1. Physical and Mine-ralogical Methods - Agronomy Monograph 9 (2^nd Edition). Pp. 383-411. Am. Soc. Agron. Inc. - Soil Sci. Soc. Am. Inc., Madison, Wisconsin, USA.
12. Gregorich, EG; MH Beare; UF McKim & JO Skjemstad. 2006. Chemical and biological characteristics of physically uncomplexed organic matter. Soil Sci. Soc. Am. J. 70: 975-985.         [ Links ]
13. Hao, X & AN Kravchenko. 2007. Management practice effects on surface soil total carbon: differences along a textural gradient. Agron. J. 99: 18-26.
14. Hassink, J. 1997. The capacity of soils to preserve organic C and N by their association with clay and silt. Plant Soil 191: 77-87.
15. Hevia, GG; DE Buschiazzo. In: Hepper; AM Urioste & EL Antón. 2003. Organic matter in size fractions of soils of the semiarid Argentina. Effects of climate, soil texture and management. Geoderma 116: 265-277.         [ Links ]
16. INTA. 1979. Carta de Suelos de la República Argentina. Hoja: 3757-31, Balcarce., Instituto Nacional de Tecnología Agropecuaria, Secretaría de Agricultura, Ganadería y Pesca, Buenos Aires, Argentina.
17. Irízar, AB. 2010. Cambios en las reservas de materia orgánica del suelo y sus fracciones granulométricas: efecto de la secuencia de cultivo, del sistema de labranza y de la fertilización nitrogenada. M.Sc. Thesis, Escuela para Graduados Alberto Soriano, Facultad de Agronomía, Universidad de Buenos Aires.
18. Janzen, HH. 2006. The soil carbon dilemma: shall we hoard it or use it? Soil Biol. Biochem. 38: 419-424.
19. Kobayashi, K & MU Salam. 2000. Comparing simulated and measured values using mean squared deviation and its components. Agron. J. 92: 345-352.         [ Links ]
20. Liang, A; X Yang; X Zhang; N McLaughlin; Y Shen & W Li. 2009. Soil organic carbon changes in particle-size fractions following cultivation of Black soils in China. Soil Tillage Res. 105: 21-26.
21. Manuel-Navarrete, D; GC Gallopín; M Blanco; M Díaz-Zorita; DO Ferraro; H Herzer; P Laterra; MR Murmis; G Podestá; J Rabinovich; EH Satorre; F Torres & EF Viglizzo. 2009. Multi-causal and integrated assessment of sustainability: the case of agriculturization in the Argentine Pampas. Environ. Dev. Sustain. 11: 621-638.
22. MAGyP. 2011. Estadísticas de producción agrícola. Ministerio de Agricultura, Ganadería y Pesca, Buenos Aires, Argentina. www.sagpya.mecon.gov.ar. Last access October 15, 2011.
23. McConkey, BG; BC, Liang; CA, Campbell; D, Curtin; A, Moulin; SA, Brandt & GP, Lafond. 2003. Crop rotation and tillage impact on carbon sequestration in Canadian prairie soils. Soil Tillage Res. 74: 81-90.         [ Links ]
24. McLauchlan, KK. 2006. Effects of soil texture on soil carbon and nitrogen dynamics after cessation of agriculture. Geoderma 136: 289-299.
25. Percival, HJ; RL Parfitt & NA Scott. 2000. Factors controlling soil carbon levels in New Zealand grasslands: is clay content important? Soil Sci. Soc. Am. J. 64: 1623-1630.         [ Links ]
26. Plante, AF; RT Conant; CE Stewart; K Paustian & J Six. 2006. Impact of soil texture on the distribution of soil organic matter in physical and chemical fractions. Soil Sci. Soc. Am. J. 70: 287-296.
27. Puget, P & R Lal. 2005. Soil organic carbon and nitrogen in a Mollisol in central Ohio as affected by tillage and land use. Soil Tillage Res. 80: 201-213.         [ Links ]
28. Quiroga, A; D Funaro; E Noellemeyer & N Peinemann. 2006. Barley yield response to soil organic matter and texture in the Pampas of Argentina. Soil Tillage Res. 90: 63-68.         [ Links ]
29. Quiroga, A; R Fernández; D Riestra; E Noellemeyer & N Peinemann. 2008. Influencia del manejo sobre la distribución del tamaño y estabilidad de agregados en suelos influenciados por variaciones en la granulometría y contenidos de materia orgánica. Proceedings 21st Congreso Argentino de la Ciencia del Suelo, May 13-16, 2008, Salta-Jujuy, Argentina, 6 p. (in CD-ROM).
30. R Development Core Team. 2009. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. www.R-project.org. Last access June 1, 2011.
31. Roldán, MF. 2011. Distribución del carbono orgánico particulado por tamaño de agregados bajo distintos sistemas de labranza. M.Sc. Thesis, Facultad de Ciencias Agrarias, Universidad Nacional de Mar del Plata.
32. Schlichting E; HP Blume & K Stahr. 1995. Bodenkundliches Praktikum. Paul Parey. Hamburg, Germany.
33. Sainz Rozas, HR; HE Echeverría & HP Angelini. 2011. Niveles de materia orgánica y pH en suelos agrícolas de la Región Pampeana y Extrapampeana. Ciencia del Suelo 29: 29-37.         [ Links ]
34. Six, J; RT Conant; EA Paul & K Paustian. 2002. Stabilization mechanisms of soil organic matter: Implications for C-saturation of soils. Plant Soil 241: 155-176.         [ Links ]
35. Six, J; H Bossuyt; S Degryse & K Denef. 2004. A history of research on the link between (micro) aggregates, soil biota, and soil organic matter dynamics. Soil Tillage Res. 79: 7-31.         [ Links ]
36. Studdert, GA & HE Echeverría. 2000. Crop rotations and nitrogen fertilization to manage soil organic carbon dynamics. Soil Sci. Soc. Am. J. 64: 1496-1503.         [ Links ]
37. Studdert, GA; HE Echeverría & EM Casanovas. 1997. Crop-pasture rotation for sustaining the quality and productivity of a Typic argiudoll. Soil Sci. Soc. Am. J. 61: 1466-1472.         [ Links ]
38. USDA. 2006. Keys to Soil Taxonomy, 10th ed., Washington, DC, USA.
39. Varvel, GE & WW Wilhelm. 2010. Long-term soil organic carbon as affected by tillage and cropping systems. Soil Sci. Soc. Am. J. 74: 915-921.         [ Links ]
40. Zinn, YL; R Lal; JM Bigham & DVS Resck. 2007. Edaphic controls on soil Organic Carbon retention in the Brazilian Cerrado: texture and mineralogy. Soil Sci. Soc. Am. J. 71: 1204-1214.         [ Links ]