SMALL RELIEF SHAPE VARIATIONS INFLUENCE SPATIAL VARIABILITY OF SOIL CHEMICAL ATTRIBUTES

Soils with small variations in relief and under the same management system present differentiated spatial variabilities of their attributes. This variability is a function of soil position in the landscape, even if the relief has little expression. The aim of this work was to investigate the effects of relief shape and depth on spatial variability of soil chemical attributes in a Typic Hapludox cultivated with sugar cane at two landscape compartments. Soil samples were collected in the intercrossing points of a grid, in the traffic line, at 0-0.2 m and 0.6-0.8 m depths, comprising a set of 100 georeferenced points. The spatial variabilities of pH, P, K, Ca, Mg, cation exchange capacity and base saturation were quantified. Small relief shape variations lead to differentiated variability in soil chemical attributes as indicated by the dependence on pedoform found for chemical attributes at both 0-0.2 m and 0.6-0.8 m depths. Because of the higher variability, it is advisable to collect large number of samples in areas with concave and convex shapes. Combining relief shapes and geostatistics allows the determination of areas with different spatial variability for soil chemical attributes.


INTRODUCTION
Scientific experimentation assumes the existence of random variability for soil attributes.Nevertheless, several studies have demonstrated that soil physical and chemical attributes are correlated with soil spatial distribution (Yost et al., 1982;Vieira et al., 1983;Tsegaye & Hill, 1998;Carvalho et al., 2003).The distribution of these attributes may be influenced by successive soil management which leads to variability, even in soils considered highly homogenous, e.g.Oxisols that cover a large area of cultivated soils i n Brazil (Buol, 1990).
Working with landscape position can contribute to studies of soil chemical attributes (Kravchenko & Bollock, 2000;Marques Jr. & Lepsch, 2000;Mohanty & Mousli, 2000;Pachepsky et al., 2001).Landscape variability associated with topographic features affects spatial patterns of water movement on both soil surface and subsurface, thus being able to affect nutrient availability (Li et al., 2001).It is supposed that relief (Halbfass & Grunewald, 2003) and microrelief (Kuzyakova et al., 1997) are also factors that greatly affect soil spatial variability.Concave and convex relief shapes show different variability as compared with linear micro-relieves, mainly because of the water flow.The aim of this work was to investigate the effects of small variations in relief shape and depth on the spatial variability of soil chemical attributes in a Typic Hapludox cultivated with sugar cane at two landscape compartments.

MATERIAL AND METHODS
The studied site was located in Guariba, SP, Brazil (21º19'S, 48º13'W; altitude 640 m).The area's climate is Cwa according to Köppen's classification.The mean annual rainfall is 1.400 mm; precipitation is concentrated from November to February.
The area's relief is soft wavy and the slope varies from 3 to 8 %.The site has been cropped with sugar cane for 30 years and it was the 5th harvest at the moment of soil sampling.The Typic Hapludox, a clay-textured soil (LVef), was originated from Basalts of São Bento Group, "Serra Geral" formation.
According to the Troeh's model (Troeh, 1965), the curvature and profile of the landscape in the lower 1/3 of the hillslope were classified into two compartments (I and II; Figure 1).In the compartment I the relief was basically linear-shaped, while in compart-ment II it was concave-and convex-shaped.Soil samples were collected in the intercrossing points of a grid (10 × 10 m; area = 1 ha), in the traffic line and at 0-0.2 m and 0.6-0.8m depths.One hundred points were georeferenced in the grid.
The active acidity (pH in CaCl 2 0.01 moL L -1 ) was determined by a potentiometer using a 1:2.5 soil:CaCl 2 ratio, and the potential acidity (H+Al) was determined according to Raij et al. (2001).The ion exchange resin method (Raij et al., 2001) was used for Ca, Mg and K and available P extractions.Based on chemical analyses, cation exchange capacity (CEC) and base saturation (%BS) were calculated.
Sampling size was calculated by the Equation (1) described by Warrick & Nielsen (1980): where N is the minimal number of samples; T α is the value of Student's T test at 95 % probability level; σ is the coefficient of variation; and D is the percentage of variation respective to the mean (5, 10 or 15 %).
Using the regionalized variable theory, the spatial variation was calculated by the semiovariogram method (Journel & Huijbregts, 1991) that assumes the stationarity of the intrinsic hypothesis and is expressed by the following equation: which is a function of the vector h, and thus depends on the magnitude and direction of h.The semivariance was estimated by the following equation: where N(h) is the number of measured point pairs Z(x i ), Z(x i + h), separated by a vector h.= C 0 + C 1 [1 -exp (-3h/a)], for 0 < h < d, where C 0 is the nugget effect, C 0 + C 1 is the sill, a is the range and d is the maximal distance in which the semiovariogram is defined.To determine the spatial dependence, the semivariogram examination was performed with the program GS + (Gamma Design Inc., Plainwell, MI).When more than one variogram could be used, the most appropriated was chosen by the cross-validation method.The software Surfer 7.0 (Golden Software, Inc., New York, NY, 1999) was used for plotting the spatial distribution maps of the variables.The spatial dependence of the investigated variables was performed according to the classification of Cambardella et al. (1994).

RESULTS AND DISCUSSION
The minimal number of samples required with 5, 10 and 15 % of variation respective to the mean was assessed for each compartment by the Warrick & Nielsen's formula (Table 1) (Warrick & Nielsen, 1980).The required number of samples was relatively high for P, K, Ca, Mg, H+Al, CEC and BS%, as also described by Silveira et al. (2000).On the other hand, a small number of sub-samples was necessary for estimating pH, in agreement with Tsegaye & Hill (1998) and Silveira et al. (2000).Generally, a higher number of samples was required in the compartment II where all studied variables presented higher variability.Similar results were found by Montanari et al. (2005) for Table 1 -Estimative of the minimal number of samples required with 5, 10 and 15% variation respective to the mean in each landscape compartment.
the spatial variability of chemical attributes in soil cropped with sugar cane in different pedoforms.This shows that small variations in relief must be considered for sampling delineations in experiments with the same scale.
Data distribution was normal for the CEC and %BS, in both depths, and for the K, in the 0.0-0.2m depth, according to the Kolmogorov-Smirnov's test (α = 0.05) (Table 2).Dobermann & George (1994) found normal distribution of pH, and Souza et al. (2003) to P, K and %BS in a study of the effect of relief on anisotropic variation of chemical attributes at 0-0.2 m depth.Despite showing some asymmetric distributions, mean and median values were similar for all the investigated variables which indicate that the measures of central tendency are not dominated by atypical values (Cambardella et al., 1994).However, any theoretical distribution is always matched inexactly because data from nature are under the effect of several concomitant factors (Cressie, 1991).
For H+Al, CEC and %BS, the CVs were between 12 and 24 %, higher than the CV observed for pH.Variabilities of H+Al were similar to those found in an Oxisol at the same depths here in considered (Souza et al., 2003), and in a Podzolic with native vegetation (Park & Vlek, 2002).CVs within this range were also reported in an Oxisol cropped with corn during the summer and winter, considering only the 0-0.2 m depth (Silveira et al., 2000).Considering the %BS, this range of variability was similar to those reported to an Ultisol under native vegetation at 0-0.2 m depth (Park & Vlek, 2002) and a sugar cane-cropped Oxisol at 0-0.2 m and 0.6-0.8m depths (Souza et al., 2003).
Registered CVs for CEC are similar to those described for a corn and soybean cropped Molisol (Cambardella et al., 1994), to a 30-year sugar canecropped Oxisol (Souza et al., 2003) and to a corn cropped Entisol sampled at the same soil depths investigated here (Castrignanò et al., 2000).P, K, Ca and Mg showed higher variabilities, above 24% at the two studied depths.At these same depth conditions, Souza et al. (2003) found variabilities higher than 24 %.This is similar to data reported to a corn and soybean cropped Molisol (Cambardella et al., 1994)  depth (Carvalho et al., 2003).High CVs were reported for Ca and Mg, and extremely high values were reported for P (Chien et al., 1997); Castrignanò et al. (2000) found high Ca and Mg CV values for corncropped Typic Xerofluvent.Moderate CV is not necessarily a good indicator of spatial variability of soil attributes because of differences in soil nutrients caused by management procedures (Wollenhaupt et al., 1997).
The Oxisol concept implies that soil characteristics are homogeneous, both in profile and horizontal extension.Nevertheless, a 30-year sugar cane-cropped surface was not homogeneous even in the surface, as showed here by the high variability of soil attributes, especially K, P, Ca and Mg.Several factors affect the extension of the spatial variability in a soil unity, as well as the degree of this variability extension affects the association between productivity and soil surveys (Sadler & Russel, 1997).
Results for the geostatistics analysis are presented in Table 3.All the analyzed variables presented spatial dependence for both investigated depths.K, pH, Ca, H+Al, CEC and %BS fitted to a spherical model and P and Mg to an exponential model at 0-0.2 m depth.However, at 0.6-0.8m depth, only K, Ca, H+Al and CEC fitted better to a spherical model, while pH, P, Mg and %BS fitted better to an exponential model.
In a study of the effect of intensive soil use on plant and soil nutrient spatial variability, it was found that all these variables fitted well to the spherical model (Tsegaye & Hill, 1998).In another work soil attributes also fitted well to a spherical model (Cambardella et al., 1994).
The nugget effect (Table 3) represents the random variation usually derived from the inaccuracy of measurements or variations of the properties that cannot be detected in the sample range (Trangmar et al., 1985).The nugget effect in the semiovariograms found here were not high because of the short distance among consecutive sampling points.The parameters of the semiovariogram (sill and range) for the soil chemical attributes are similar between depths, without significant modifications in the spatial dependence (Table 3).The relation C 0 /(C 0 +C 1 ) indicates that the investigated variables showed a moderate spatial dependence, except for Ca that expressed a stronger spatial dependence at 0-0.2 m depth.At the 0.6-0.8m depth, pH, P, Mg and %BS showed moderate spatial dependence, while K, Ca, H+Al and CEC were strongly dependent on the soil's spatial distribution.The higher C 0 /(C 0 +C 1 ) values indicate that the soil may present lower-scaled variability than those presented here, and that more intensive sampling may reveal a higher spatial continuum of the attributes analyzed.The higher spatial dependence of chemical attributes occurred at 0.6-0.8m depth and then was associated to the low level of management.The spatial variability of soil attributes may be affected by intrinsic and extrinsic factors, the latter resulting usually from soil management (Cambardella et al., 1994).
The variable range can be sequenced in increasing order for the 0-0.2m depth, as follow (values in Table 3): P < Ca < K < %BS < pH < H+Al < CEC < Mg.The following sequence was found for the 0.6-0.8m depth: H+Al < Ca < K < CEC < Mg < P < %BS < pH.These values were always lower at the 0.0-0.8m depth, except for pH and P, indicating a higher spatial distribution discontinuity for the soil chemical attributes at higher depth.Considering P, the higher value at the surface layer (0-0.2 m) can be a function of fertilization in this layer, because this nutrient has a low mobility in the soil (Raij et al., 2001).For the pH, the sequence is probably also determined by effects of nitrogenous fertilization.These results are similar to those reported by Souza et al. (2003) for P. Soil management with fertilizers and turning over the ground may contribute to the increase in the range and decrease spatial dependence, characterizing a higher continuum in the distribution of the elements at the surface sampled layer.
The chemical attributes and Kriging's maps vary from low to very high fertility (Table 2, Figures 2 and 3).These values show large amplitudes which imply that in some places of the area the fertilizer application may be excessive.The higher values detected in the deeper layer (0.6-0.8) in comparison to the surface, may be even due to this high amplitude.Inadequate soil chemical management hampers fertilization efficiency and cause leaching of excessive fertilizers (Bhatti et al., 1991), therefore resulting in economical losses from both nutrient unbalance and environmental impact.
In the higher region with homogeneous profile and relief shape (compartment I), the variability was lower (Figures 1, 2, and 3) for all the studied variables.On the other hand, in the lower region, in the most heterogeneous area (compartment II), the variability was higher.This relation with the landscape position has also been reported by other studies.Geostatistics may help evaluation of spatial dependence of chemical attributes in the landscape, with the nutrient availability at least partially controlled by the position of the landscape (Amador et al., 1997 andVaidya &Pal, 2002).Furthermore, Brubaker et al. (1993) and Vaidya & Pal (2002) reported that soil chemical attributes and position in the landscape are interrelated, as also reported in the present study.
Differences in spatial variability of soil chemical attributes in different pedoforms are probably associated to small variations in relief shape, which was the principal factor determining the highest variability found in the concave and convex pedoforms.Similar results were found by Souza et al. (2003;2004).These results support the idea that lateral and superficial fluxes of water in concave and convex forms determine specific environments that interfere in pedogenetic processes and favor the development of a higher spatial variability for soil attributes (Franzen et al., 1998).

Table 2 -
Descriptive (mean, median, coefficient of variation (CV), asymmetry and kurtosis) and normality analyses of soil attributes.