SPATIAL CONTINUITY OF ELECTRICAL CONDUCTIVITY , SOIL WATER CONTENT AND TEXTURE ON A CULTIVATED AREA WITH CANE SUGAR

Spatial variability of soil attributes affects crop development. Thus, information on its variability assists in soil and plant integrated management systems. The objective of this study was to assess the spatial variability of the soil apparent electrical conductivity (ECa), electrical conductivity of the saturation extract (ECse), water content in the soil (θ) and soil texture (clay, silt and sand) of a sugarcane crop area in the State of Pernambuco, Brazil. The study area had about 6.5 ha and its soil was classified as orthic Humiluvic Spodosol. Ninety soil samples were randomly collected and evaluated. The attributes assessed were soil apparent electrical conductivity (ECa) measured by electromagnetic induction with vertical dipole (ECa-V) in the soil layer 0.0.4 and horizontal dipole (ECa-H) in the soil layer 0.0-1.5 m; and ECse, θ and texture in the soil layers 0.0-0.2 m and 0.2-0.4 m. Spatial variability of the ECa was affected by the area relief, and had no direct correlation with the electrical conductivity of the saturation extract (ECse). The results showed overestimated mean frequency distribution, with means distant from the mode and median. The area relief affected the spatial variability maps of ECa-V, ECa-H, ECse and θ, however, the correlation matrix did not show a well-defined cause-and-effect relationship. Spatial variability of texture attributes (clay, site and sand) was high, presenting pure nugget effect.


INTRODUCTION
Precision agriculture requires determination and analysis of spatial and temporal variations of production factors, especially of the soil.These studies assist in determining specific management sites (SIQUEIRA; SILVA; DAFONTE, 2015;SIQUEIRA et al., 2016a), enabling variable rate input applications and determination of appropriate time of application, thus increasing crop yield (SILVA et al., 2013).
Thematic maps are among the main tools used to assess factors affecting crop development.Maps are used in precision agriculture to manage spatial and temporal variability of crop factors, guiding specific agricultural practices to improve efficiency of input application, reducing production costs, impacts on the environment (MOLIN; RABELO, 2011;GUO;MAAS;BRONSON, 2012;ALVES et al., 2013), and soil compaction caused by machinery traffic.
Shaner, Farahani and Buchleiter (2008) also emphasized the importance of EC a to determine sites for specific soil management, due to its correlation to different soil physical and chemical attributes that affect crop yield.
The determination of the EC a measured by electromagnetic induction is related to different soil properties because its readings are the result of the interactions between soil porous spaces, which are filled with air or water, interactions between soil particles, and structure state (SIQUEIRA; SILVA; DAFONTE, 2015;SIQUEIRA et al., 2016a).Thus, information on the correlations of EC a measured by electromagnetic induction to other soil properties in different types of soil and crops is important (SIQUEIRA; SILVA; DAFONTE, 2015;SIQUEIRA et al., 2016b).
Electromagnetic induction is an important alternative to evaluate EC a , since it is a noninvasive technique that evaluate EC a in the soil profile through multiple readings (ABDU; ROBINSON;JONES, 2007).
The objective of this study was to assess the spatial variability of the soil apparent electrical conductivity (EC a ), electrical conductivity of the saturation extract (EC se ), water content in the soil (θ % ) and soil texture (clay, silt and sand) of a sugarcane crop area in the State of Pernambuco, Brazil.

MATERIAL AND METHODS
The experiment was carried out in an area of about 6.5 ha of the Santa Teresa sugar and alcohol industry, in Goiana, Zona da Mata Norte, State of Pernambuco, Brazil (07°34'25''S, 34°55'39''W and average altitude of 8.5 m) (Figure 1).(2013).Textural classification of the soil (Table 1) was determined using the methodology recommended by EMBRAPA (2011).
The climate of the region is tropical humid type As', i.e., hot and humid, according to the classification of Köppen, with a rainy season from autumn to winter, annual average precipitation of 1,924 mm and annual average temperatures of 24 °C.
The study area has been used for rainfed sugarcane (Saccharum officinarum L.) crops, grown as single-crop, with straw burning before harvesting, since 1988.The crop area had been renewed in the 2010-2011 crops season; the soil was plowed, harrowed, grooved, limed, and fertilized and the sugarcane variety RB867515 was planted.
Ninety sampling points were randomly chosen in the study area (Figure 2) and georeferenced with a GPS device with differential correction to subsequent data collection of the soil texture (clay, silt and sand), electrical conductivity of the saturation extract (EC se ) and water content.Samplings were carried out in January 21, 2014, with texture, EC se and water content determined in the soil layers of 0.0-0.2 and 0.2 -0.4 m.Field evaluations of soil apparent electrical conductivity (EC a ) (mS m -1 ) was carried out using an electromagnetic induction device (EM38) (GEONICS, 1999), which measures the horizontal dipole (EC a -H), with readings within the soil layer 0.0-0.4m and the vertical dipole (EC a -V), with readings within the layer 0.0-1.5 m, following the procedures described by Siqueira, Silva and Dafonte (2015) and Siqueira et al. (2016b).
Field evaluations of the volumetric water content in the soil (θ % ) in the soil layers 0.0-0.2 and 0.2-0.4m was carried out using a transmission line oscillator (Hydrosense ® , Campbell Scientific Australia Pty. Ltd.), which has a probe that emits an electromagnetic signal in the soil and evaluates how many times the signal returns in a certain period of time (SIQUEIRA et al., 2015).
Laboratory evaluations of the soil texture (clay, silt and sand) (g kg -1 ) and EC se (dS m -1 ) were carried out in samples of the soil layers 0.0-0.2 and 0.2-0.4 m.The samples were air dried, disaggregated, sieved in a 2 mm mesh sieve.Soil texture (g kg -1 ) was determined with a densimeter and EC se by the saturated paste extract method, following the procedures described by EMBRAPA (2011).
The means of the attributes were subjected to the main statistical procedures (mean, median, standard deviation, coefficient of variation, skewness and kurtosis).The normality of the data was evaluated through coefficients of skewness and kurtosis and histograms of frequency distribution.The coefficient of variation (CV, %) was classified as low (<12%), intermediate (12% to 62%) and high (>62%) (WARRICK; NIELSEN, 1980).The linear correlation between the attributes was determined with significance level of 1% using the Shapiro-Wilk test, including the relief data of all sampling points to assess the effect of relief on the variables.Statistical analyzes were performed using software R 3.3.1 (R CORE TEAM, 2016).
Spatial dependence analysis was performed by adjusting of the experimental semivariogram, based on the assumption of stationarity of the intrinsic hypothesis (VIEIRA, 2000;SIQUEIRA et al., 2015).Spatial autocorrelation between neighboring sampling points was calculated by the semivariance γ(h), which is estimated by the Equation (1), in which N (h) is the number of experimental pairs of observations Z (x i ) and Z(x i + h) separated by the distance h.
The software Surfer 11.0 was used to develop maps of spatial variability.Isoline maps were developed when the pure nugget effect was detected to compare the attributes, using the Surfer's default parameters, which is based on a linear interpolation model by kriging.

RESULTS AND DISCUSSION
According to the mean and median analysis (Table 2), the data of all variables tended to normality.However, the analysis of frequency distribution graphs (Figures 3 and 4) showed different distributions (symmetrical and asymmetrical).The coefficients of skewness and kurtosis were different than 0 and 3, thus, the data did not show normal distribution.EC a -V = soil apparent electrical conductivity measured by electromagnetic induction with vertical dipole in the soil layer 0.0-0.4,EC a -H = soil apparent electrical conductivity measured by electromagnetic induction with horizontal dipole in the soil layer 0.0-1.5 m, EC se = electrical conductivity of the saturation extract, SD = standard deviation; CV = coefficient of variation (%).The means of EC a -V and EC a -H were different.According to Siqueira, Silva and Dafonte (2015) and Siqueira et al. (2016a), the largest differences in soil apparent electrical conductivity, measured by electromagnetic induction (EC a -V and EC a -H) are due to soil relief, water rate fluctuation, water content, texture and organic matter content.The water content in the soil and soil texture varied in both depths, explaining the greatest differences between EC a -V and EC a -H.Moreover, 80% of the EC a -V readings were directly related to the EC a -H readings, as also found by Geonics (1999) and Siqueira, Silva and Dafonte (2015).Therefore, despite the different means of EC a -V and EC a -H, these variables are correlated.
EC a -V (mS m -1 ), EC a -H (dS m -1 ) and EC se (dS m -1 ) means were different.However, despite representing the same soil attribute, they were evaluated through different methods and expressed in different scales.Their major differences were due to evaluation method, since the electromagnetic induction method is assessed in the field, considering the soil electric current flow as a three-dimensional body, encompassing a larger volume of soil (consisting of porous spaces, water and mineral particles), and EC se is determined in laboratory, under controlled conditions, using disturbed soil samples, with readings that consider only the salts of the soil solution.(SIQUEIRA et al., 2014;SIQUEIRA;SILVA;DAFONTE, 2015).
The coefficient of variation (CV%) of clay and sand was classified as low (< 12%); water content in the soil (θ % ) and EC se had intermediate CV (12% to 62%); and EC a -V, EC a -H and silt content had high CV (> 62%).
According to the frequency distribution histograms (Figure 3), most attributes had lognormal distribution, however, geostatistical analysis can be carried out despite the data normality (VIEIRA, 2000).
The frequency distribution graphs for EC a -V and EC a -H showed leptokurtic positively skewed distribution, i.e., there were many low EC a -V and EC a -H, thus, their mode and median were close and their means were overestimated.The EC se of the soil layer 0.0-0.2m also had leptokurtic positively skewed distribution, whereas the EC se of the soil layer 0.2-0.4m had normal frequency distribution, with slightly trend to a negatively skewed distribution.The histograms for EC a -V, EC a -H and EC se was probably affected by the relief, as reported by Siqueira, Silva and Dafonte (2015), who found the relief affecting the water flow in the soil and consequently, the EC a -V, EC a -H and EC se .
The water content in the soil (θ % ) had lognormal frequency distribution, also with overestimation of the mean and leptokurtic positively skewed distribution.This result was expected, since the water flow and distribution in the soil favor the formation of sites with high and low water content as a function of relief, as reported by Siqueira et al. (2015).The frequency distribution histograms for water content in the soil showed very elongated tails, confirming that the water content varied, showing areas of high and low water content along the landscape of the study area.
Only the data of silt, from the texture attribute, had lognormal frequency distribution in both soil layers.Data of clay and sand had normal distribution, with more homogeneous histograms and less elongated tails, resulting in more stable means.
According to the geostatistical analysis (Table 3), most of the texture attributes had pure nugget effect (PNE), denoting a small scale spatial variability, i.e., at distances smaller than that chosen by random sampling.Only the model for clay content of the soil layer 0.2-0.4m fitted to the experimental semivariogram.
The spherical model was fitted to the semivariograms of EC a -V, EC a -H and θ (0.0-0.2); and the Gaussian model to EC se in both layers.The spherical model fit to the semivariograms for most of the attributes, confirming reports of other authors, who describe this model as that that best fit to the attributes of the soil (CAMBARDELLA et al., 1994;VIEIRA, 2000;SIQUEIRA;SILVA;DAFONTE, 2015;SIQUEIRA et al., 2016a).
The highest range (a) (m) was found for the EC se in the soil layer 0.2-0.4m (199 m) and the lowest, for the EC a -H in the soil layer 0.2-0.4m (57 m).
According to the classification of Cambardella et al. (1994), the attributes evaluated had strong (< 25%,) and moderate (25 to 75%) spatial dependence index.Siqueira et al. (2015) evaluated the spatial variability of soil attributes with different scales and found high SDI (%) for water content in the soil (%) at different soil depths (0.0-0.2, 0.2-0.4 and 0.4-0.6 m).Differences in spatial dependence index were due to the soil natural variation and relief of the study area.
Parameters of models fitted to the experimental semivariogram of EC a -V and EC a -H showed a similar spatial pattern, fitting a spherical model.The EC se spatial pattern was different, especially by fitting a Gaussian mathematical model.This result was due to the scalar magnitude and because readings were performed in undisturbed (EC a -V and EC a -H) and disturbed (EC se ) soil samples.
According to the linear correlation matrix (Table 4), the relief was significantly correlated at 1% of probability (Shapiro-Wilk test) only to EC a -V (| r | = 0,815) and to EC a -H (r = 0.826).
According to the spatial variability maps (Figures 5 and 6), the EC a -V (Figure 5A) and EC a -H (Figure 5B) had similar distribution of the contour lines, explaining their high correlation (|r| = 0.940).Moreover, the device used reads a same volume of soil, and vertical dipole readings (EC a -V) are affected by the soil surface layer, which was evaluated by the horizontal dipole (EC a -V) (CORWIN;LESCH, 2003LESCH, , 2005;;SIQUEIRA;SILVA;DAFONTE, 2015).
The spatial variability maps of EC a -V, EC a -H, EC se and θ showed no similar patterns (Figure 5), confirming their low spatial correlation (Table 4).However, these maps followed a same trend pattern, as shown in the relief map (Figure 1).Therefore, the spatial distribution of the attributes (EC a -V, EC a -H, EC se and θ) is affected by relief.According to Siqueira, Silva and Dafonte (2015) and Siqueira et al. (2015), soil declivity is the factor that most affect water distribution and consequently, the distribution and interaction of other soil attributes.
Spatial variability maps of texture (clay, silt and sand) (Figure 6) in the soil layers 0.0-0.2 and 0.2-0.4m (Figure 6) showed no spatial relationship with the maps of EC a -V, EC a -H, EC se and θ (Figure 5), confirmed by the low values of linear correlation (Table 4).
The spatial distribution maps of soil texture (clay, silt and sand) showed great difference in contour lines, denoting high spatial variability.All texture attributes had pure nugget effect (PNE), except the clay at 0.2-0.4m (Table 3).These maps were developed by linear interpolation to compare spatial patterns, even with PNE, since cartography is a classical science, and data with PNE processed by geostatistics are usually not properly analyzed.Thus, the PNE of texture attributes was due to the high variability of the data along the landscape, affected by different soil formation factors (SIQUEIRA;SILVA;DAFONTE, 2015;SIQUEIRA et al., 2015).

CONCLUSIONS
Spatial variability of the soil apparent electrical conductivity measured by electromagnetic induction (EC a -V and EC a -H) was affected by relief and had no direct correlation to the electrical conductivity of the soil saturation extract (EC se ).
The soil attributes evaluated had frequency of distribution with overestimated means, and means distant from the mode and median.
The area relief affected the spatial variability of EC a -V, EC a -H, EC se and θ, however, the correlation matrix did not show a well-defined causeand-effect relationship.
Spatial variability of soil texture attributes (clay, site and sand) was high, presenting pure nugget effect.

Figure 1 .
Figure 1.Topographic map of the study area.

Figure 3 .
Figure 3. Histograms of frequency distribution of the soil attributes evaluated.

Figure 4 .
Figure 4. Histograms of frequency distribution of the soil attributes evaluated.
1 Figure 2. Location of the sampling points in the study area.

Table 2 .
Descriptive statistics of attributes of an orthic Humiluvic Spodosol of sandy texture.