Spatial Variability in Stability of Aggregates and Organic Matter of an Oxisol

Soil preparation may break its structure, destabilize the aggregates, and cause the loss of organic matter (OM). The study of spatial variability of soil attributes is an important indicator of soil physical quality. The aim of this study was to describe the spatial variability of the stability of aggregates and organic matter in Oxisol (Yellow Latosol) under different management systems.We collected simple samplings of soil in the eastern mesoregion of Maranhão, Brazil. Experimental areas with two distinct management systems were studied: conventional tillage and no-tillage. In each experimental area, we fitted a rectangular mesh of 50 points with 40m of spacing and 0.00 to 0.20 mof depth. The response variables were: weighted mean diameter (WMD); geometric mean diameter (GMD); percentage of aggregates (on classes of size between 1-2 mm and above 2 mm); and organic matter (OM). The no-tillage management showed high values of WMD, GMD, class of aggregates and OM. Maps of WMD and GMD were spatially correlated to OM map at no-tillage management. Soil properties had a spatial-dependent structure. The management system influenced the stability of aggregates and the amount of organic matter.


Introduction
Soil preparation breaks its structure, which destabilizes the aggregates and leads to losses of organic matter.The stability of aggregates is an important indicator to assess the physical quality of the soil.The aggregation influences gas exchange, infiltration, retention, and water availability.It also affects soil resistance to compaction (Braida, Bayer, Arbuquerque, & Reichert, 2011), which associated with good land use and management practices, contributes to raising soil organic matter (OM), an essential condition for high productivity (Dexter & Youngs, 1992).
The study of spatial variability using geostatistical techniques has been widely used to evaluate changes in the behavior of physical and chemical attributes of soil (Coutinho Alho, Costa Campos, Pinheiro da Silva, Campos Mantovanelli, & Menezes de Souza, 2014;Corado Neto et al., 2015;Ribeiro et al., 2016).Analyses of semivariance and ordinary kriging interpolation allow the estimation of studied attributes in non-sampled areas, besides being a valuable tool for plans of further samplings, management of specific sectors, and optimization of agriculture inputs.
At the grain production region of Brasilian savanna, farmers have difficult to maintaining plant cover during the off-season of soybean and maize.The mechanized soil technologies and torrential rains at the beginning of the plantations causes large losses of sediment and consequent water erosions.Using geostatistical technics might help to improve the management practices and optimize the crop.At this context, we aim to characterize the spatial variability of the stability of aggregates and organic matter, using geostatistical techicnics, in Oxisoil (Yellow Latosol) under different management systems.jas.ccsenet.

Materia
The To identify the spatial variability of soil attribute, we used the semivariogram method following the principles of the intrinsic hypothesis (Isaaks & Srivastava 1989).We calculated the variance by the separation distance between samples to determine the experimental semivariograma according to the following Equation 2 where, represents the experimental semivariance for a distance of h; z(x i ) is the value of the attribute for the sample point I; and N(h) is the number of pairs of points separated by distance h.The parameters for the theoretical model of the semivariogramwere defined as: nugget; structural variance; sill; and range.Nugget is the value of semivariance for the distance zero and is the random component of variance.Sill is the value ofsemivariance at which the curve stabilizes at a constant value.The range is the distance from the origin to where the plateau reaches stable values, showing the distances beyond which samples are not correlated.
We choose the best model comparing the correlation coefficient obtained by the cross-validation technique and the highest coefficient of determination (R 2 ).Models with a coefficient of determination close to 1 were chosen to characterize the studied phenomenon.The classification of the degree of spatial dependence (DSD) was based on the ratio between the nugget effect and sill .The spatial dependence was considered weak for DSD > 75%, moderate for DSD between 25% and 75%, and strong for DSD < 25% (Cambardella et al., 1994).
The ordinary kriging technique was used to interpolate values at non-sampled locations.This technique is based on a weighted moving average of the neighboring points, obtained by the equation: where, represents the value estimated at point 0; N is the number of values used in the estimation; λ is the weight associated with each observed value; and z(x i ) is the value observed at point i.The weights (λ i ) of each neighbor point are set using the adjusted semivariogram model, resulting in a minimum estimate of variance (Soares, 2006).

Results and Discussion
Although the values of asymmetry and kurtosis indicate some asymmetric distributions, the averages and medians of all assessed attributes are close, pointing out that the data do not show marked asymmetry (Table 1).Furthermore, the adjustment of the theoretical distribution of data found in nature is only approximate (Warrick & Nielsen, 1980).All variables at no-tillage system had a normal distribution (Table 1).Although the normality is not a requirement for geostatistical analysis, it is desirable that the distribution present little asymmetry.Distribution of data with very long tailsimpairs the analysis, especially when estimates are based on average, as the ordinary kriging technique (Isaaks & Srivastava, 1989).
The highest coefficient of variation (CV) was found for the aggregate class larger than 2 mm (56.55%), at the no-tillage management, and the lowest was obtained by OM (6.55%; classified as low variability) in the conventional management (Table 1).The coefficient of variation of GMD, WMD, and aggregate class showed moderate in variability.Although the CV is a dimensionless measure that allows comparing the variability among different attributes, it does not allow analyzing the spatial variability of soil attributes and its spatial pattern.
Table 1.Descriptive statistics of weighted mean diameter (WMD), geometric mean diameter (GMD), percentage of aggregates on class size > 2 mm and 2-1 mm, and organic matter (OM) for conventional and no-tillage managements of soybean at Oxisoil In both management systems, all attributes showed spatial dependence and adjusted to the spherical model (Table 2).The spherical model has been the most used to describe the behavior of soil attributes.Regardless of the management system, the evaluated variables showed a moderate degree of spatial dependence (DSD), according to the classification of Cambardella et al. (1994).The stronger the degree of spatial dependence, the more influenced by formation factors (texture, mineralogy) the soil attributes will be, while attributes with weak spatial dependence are more influenced by extrinsic properties, such as management.
The range represents the distance at which the sampling points are spatially correlated with each other, that is, points located in an area of radius equal to the range are more homogeneous with each other.Therefore, high ranges mean low data variability.Range estimates also might indicate sampling density.Studied attributes showed large range at no-tillage management, meaning lower variability (Table 2).The low range (high variability) of soil attributes at the conventional management might be due to the soil disturbing by the tillage.Note.DSD = degree of spatial dependence; R 2 = coefficient of determination.
The WMD, GMD, class of aggregates > 2 mm and 2-1 mm, and OM showed higher values in the soil with no-tillage than conventional management (Figure 2).The conservation tillage systems provide low disturbance to the ground and retain crop residues on the surface, which preserve its structure, retain more water in the jas.ccsenet.
surface lay feature exp showed co matter for Figure 2. percent

Conclus
The studie of aggrega

Acknowle
CV lower than 12% suggests a low variance, a CV between 12% and 60% is moderate, and a CV above of 60% suggests a high variance.

Table 2 .
Models and parameters estimated from semivariogram experiments of weighted mean diameter (WMD), geometric mean diameter (GMD), percentage of aggregates on class size > 2 mm and 2-1 mm, and organic matter (OM) for conventional and no-tillage managements of soybean at Oxisoil