THE TEMPORAL STABILITY OF THE VARIABILITY IN APPARENT SOIL ELECTRICAL CONDUCTIVITY ESTABILIDADE TEMPORAL

Apparent soil electrical conductivity (ECa) measurements can be used for crop management in precision agriculture. However, ECa is a soil attribute that presents spatial and temporal variability. It is affected by a group of factors that act simultaneously on the soil, such as the soil texture, moisture content, organic matter content and ionic concentrations in the soil solution, which complicates analysis. For soil and crop management, it is important to determine whether the pattern of the ECa spatial variability changes over time. Thus, ECa measurements have the potential for delimiting management zones that are stable over time. The objective of this work was to determine whether the spatial variability pattern of ECa is maintained over time and under different soil conditions. To this end, the ECa was measured at different soil depths using a portable sensor on two crop fields. The first step was to measure and generate an ECa map for each area. By defining a path with the maximum ECa variability, 50 sampling points were located on each field. The ECa values were measured on 20 different dates in the 0 – 20 cm, 0 – 40 cm and 0 – 60 cm soil layers. The soil water content was measured at the same points in the 0 – 20 cm layer on the same dates. The temporal stability of the ECa was analyzed using spatial and temporal variability maps, a correlation analysis and a coefficient of variation over time for each field. In both areas, the ECa exhibited temporal stability in the spatial pattern variability at the three evaluated depths, even though the soil water content values changed on each date. ECa determination presents an important alternative for mapping agricultural fields for crop management in precision agricultural systems.


INTRODUCTION
Precision agriculture has been intensively developed in recent years, especially due to rapid advances in information technology and Geographic Information Systems (GIS).However, the mapping of chemical, physical and physicochemical soil properties remains a problem that has not been completely resolved.A large number of sampling points are necessary to provide sufficient information to generate a precise map of soil attributes.The acquisition and analysis of these samples are expensive and time-consuming task, which in some cases renders the process unfeasible.To overcome this problem, new technologies for direct and remote sensing have been developed for the acquisition of relevant spatial information from production fields to reduce the money and time expended for soil map generation.
Among the new sensing technologies employed in precision agricultural systems, the apparent soil electrical conductivity (EC a ) has become an important tool for understanding the spatial variability of agricultural production systems.Many studies have reported the relationship of EC a with other soil attributes, including texture, water content, salinity, pH, cation exchange capacity (TERRÓN et al., 2011;HEIL;SCHMIDHALTER, 2011;ISLAM et al., 2012;SERRANO et al., 2012;VALENTE et al., 2012a) and crop yield (MANN et al., 2011;ALCÂNTARA et al., 2012).Therefore, EC a mapping may provide a rapid and effective means of identifying areas with similar features at a lower cost, allowing for needsbased management (FAROOQUE et al., 2012;VALENTE et al., 2012b).
It is important to understand how EC a spatial variability is related to other soil attributes of agronomic importance in precision agricultural systems.This statement is especially true when an EC a map is used to delimit management zones (FAROOQUE et al., 2012;VALENTE et al., 2012 ab).The temporal variability of this soil attribute is important information when using an EC a map.In applications in which management zones are generated, it is important that the pattern of EC a spatial variability does not change over time 151 The temporal stability… MEDEIROS, W. N. et al. (HARTSOCK et al., 2000;FARAHANI;BUCHLEITER, 2004;ISLAM et al., 2012).Grain-producing regions in Brazil have two crops per year, and the interval between harvesting one crop and planting the next is usually very short; therefore, the mapping of EC a must be performed within this interval irrespective of the soil condition.For this reason, the generated maps should depict similar patterns of EC a spatial variability regardless of the soil moisture and ambient conditions (HARTSOCK et al., 2000;FARAHANI;BUCHLEITER, 2004;ISLAM et al., 2012).If this condition is not met, it would be necessary to repeat the management zone definition process immediately before planting each crop, which would result in higher costs and more time required for managing the system.
A study of non-saline soils indicated that the EC a patterns were influenced more by stable soil attributes such as texture, organic matter and subsurface structure than by dynamic attributes such as soil moisture content and temperature (HARTSOCK et al., 2000;EIGENBERG et al., 2002;ISLAM et al., 2012;SERRANO et al., 2013).Therefore, according Farahani and Buchleiter (2004), although the magnitudes of the absolute values of EC a may changes in response to modifications in the soil dynamic properties, it is expected that the pattern of EC a spatial variability does not change significantly over time.
Considering the importance of understanding how EC a changes over time, the objective of this study was to evaluate the temporal stability of the spatial pattern of EC a at different soil depths using a portable EC a sensor.

Characterization of crop fields
This study was performed on two fields with different soil textures (EMBRAPA, 2006).The first field (Field 1) had an area of 10,703.72 m 2 .This field contains sandy clay loam and has a medium texture.The second field (Field 2) had an area of 14,078.68m 2 .This field contains a loamy sand soil with a sandy texture.
In a preliminary analysis, EC a maps were acquired for both fields.For this process, EC a was measured at 203 and 163 points on Field 1 and Field 2, respectively.The measurements were obtained using a portable device (Landviser® model LandMapper® ERM-02, Texas, USA), which uses the principle of electrical resistivity measured using a probe with four electrodes.The electrodes were configured in a Wenner matrix to measure the depth at 0 -20 cm, as described by Corwin and Hedrickx (2002) and Corwin and Lesch (2003).For both fields, the maps shown in Figure 1 were obtained using an ordinary kriging methodology (WEBSTER; OLIVER, 1992), by the softwares GS+, version 9 and Surfer, version 10.
After obtaining the EC a map for each field, a path with a wider range of values was drawn, and the locations of 50 experimental sampling points were defined.These points were marked by posts, constituting the sample units in the study (Figure 1).Each unit represented an area of 2 x 2 m, from which the EC a determinations were obtained on different dates.
A determination of the coordinates for each field limit and their respective sampling points was performed using a model ProMark3 Topographical GPS instrument (Magellan®).A post-processed differential correction was performed using GNSS Solutions® software provided by the GPS device manufacturer.A base station of the Brazilian Institute of Geography and Statistics (IBGE) was used for the DGPS correction, and the SIRGAS 2000 datum was used.

Determination of the apparent soil electrical conductivity
Three probes were prepared for measuring the EC a within the sample units at different depths.The electrodes were spaced at 20, 40 and 60 cm for measuring the EC a in the 0 -20 cm layer (EC a 20), 0 -40 cm layer (EC a 40) and 0 -60 cm layer (EC a 60), respectively.
Twenty measurements of the EC a 20, EC a 40 and EC a 60 were obtained for each of the fifty sampling units in each field.The measurements were performed between 10/19/2012 and 12/19/2012.No more than forty minutes was required to conduct all of the measurements in each field; this procedure minimized the effects of temperature and water content variations on the measured soil attributes.

Determination of the soil moisture content
The soil moisture content was determined using a thermogravimetric method.On each EC a determination date, twenty-five soil samples of deformed structures were collected in the 0 -20 cm layer.The samples were collected at each point having an even identification number (Figure 1) and stored in aluminum moisture content tins.The samples were weighed and dried at 105 ºC for 24 hours to determine the dry mass of the soil.

Analysis of the temporal variability and stability of the apparent soil electrical conductivity
An initial exploratory analysis of the data was conducted.The methodology proposed by Libardi et al. (1996) was used for checking the measured values by identifying each value classified as a candidate outlier.
After eliminating the outlier values, three different methodologies were used to analyze the variability and temporal stability of EC a .First, to evaluate the temporal stability of EC a , a linear correlation analysis between the original EC a values from different dates was performed, as suggested by Vachaud et al. (1985) and Kachanoski and De Jong (1988).Pearson correlation coefficients were calculated, and the significance was tested using Student's t-test at 5% probability.
To complement the data analysis, a chart of the temporal stability was created using the coefficients of variation (CV) determined for all sample units from each measurement date.This was performed using the methodology described by Li et al. (2007), Blackmore (2000), Shi et al. (2002), Xu et al. (2006) and Serrano et al. (2011).The coefficient of variation was calculated using Equation 1: (1) where = the coefficient of variation of the values measured on different dates for a soil attribute of the i th sample unit; = the value of the soil attribute from the i th sample unit measured on the t th date; and = the number of measured values of the soil attribute.Li et al. (2007) and Blackmore (2000) used a method based on the coefficient of variation for each sampling point.They suggested that if the coefficient of variation is less than 30%, then there is temporal stability; however, values greater than 30% are indicative of temporal instability of the analyzed variable.This method was also adopted here to evaluate the coefficient of variation.
Finally, to quantify the temporal variability of the EC a spatial patterns in each field, the measured values were normalized according to the method of Farahani and Buchleiter (2004) using Equation 2: (2) where = the normalized value of the apparent soil electrical conductivity (EC a ) on a 0 to 100 scale, calculated for each measurement date and each sample unit; = the EC a value to be normalized for a defined sample unit; = the maximum EC a value for a defined field and sampling date; and = the minimum EC a value for a defined field and sampling date.
A map with the normalized EC a data (EC a(norm) ) for each measurement date was elaborated using the software Microsoft Excel 2010, where each sample unit along the selected path is represented by a cell, resulting in 50 cells for each date.The color of each cell represented a range of values, with an EC a considered to be low if 0 ≤ EC a(norm) < 33, medium if 34 < EC a(norm) < 66, and high if 67 < EC a(norm) ≤ 100 (FARAHANI; BUCHLEITER, 2004).This procedure allowed for a visual analysis of the spatial distribution of the measured EC a values using different probes and an analysis of their behavior on each determination date.

Descriptive Statistics and Exploratory Data Analysis
A preliminary analysis of the EC a data identified the presence of outliers.The total EC a determinations in each field were as follows: in Field 1, 2.23% of the data were considered to be outliers; in Field 2, this percentage was 2.5%.Outliers can influence statistical parameters such as the mean, range, standard deviation and skewness of the data distribution (HOAGLIN et al., 1992).The removal of these observations from the dataset provides a statistical summary that best represents the distribution of the analyzed variable, especially with regards to the determination of the central tendency (LIBARDI et al., 1996).No outliers were found among the soil moisture values.
The mean EC a values that were determined for the three different soil layers (EC a 20, EC a 40 and EC a 60) in the fifty sampling units showed similar behavior within each field (Figure 2).Field 1 had EC a mean values ranging between 1.20 and 19.19 mS m -1 , and Field 2 had values ranging between 2.12 and 23.20 mS m -1 (Figure 2).Machado et al. (2006) reported minimum and maximum EC a values between 1.90 mS m -1 and 13.70 mS m -1 , respectively.Molin and Faulin ( 2013) obtained EC a values ranging from 0.60 to 16.60 mS m -1 in two different assessment years.Similar values have also been reported for non-saline soils (AIMRUN et al., 2007;ALCÂNTARA et al., 2012;VALENTE et al., 2012a).The mean soil moisture content values in the sampling units on the twenty measurement dates ranged from 18.52 to 29.43% in Field 1 (Figure 2A) and from 8.44 to 43.57% in Field 2 (Figure 2B).It was also observed that in Field 1, the soil moisture content exhibited lower variability with a slight decrease from sampling unit number 1 to sampling unit number 50 (Figure 2A).In Field 2, the soil moisture content was almost constant from unit number 1 to unit number 28 (Figure 2B).From unit 28 to unit 50, there was an increasing trend in the soil moisture content with an increase in variability.This variation in the soil moisture content was directly related to the textural variation within this area.According Libardi (2005), the clay fraction of the soil has a greater capacity to retain water, and thus, the variation in the water content in a field is related to the spatial variability of the clay content.These results demonstrate that the chemical and physical soil properties on Fields 1 and 2 are different and indicate high variation in Field 2

Temporal stability of the apparent soil electrical conductivity
The correlation coefficients values obtained using the measured EC a values from different dates were significant and high regardless of the soil layer that were evaluated for both fields (Table 1).This behavior indicates that there was a high temporal stability of the EC a in these areas during the analyzed time period.
When analyzing the temporal stability charts developed from coefficient of variation (CVi) values (Figure 3), Field 1 exhibited 90.67% of the data presented with a CVi of the EC a less than 30%, indicating a high temporal stability of this soil attribute (Figure 3A).When analyzing the behavior of the EC a 60 individually, only one of the 50 sample units in the area exhibited a CVi value greater than 30%.In Field 2, the EC a exhibited temporal instability (CVi > 30%) in sample units 1 to 27, and this stability was greater for EC a 40 and EC a 60 (Figure 3B).Follow-up field observations and soil analyses indicated that these sample units had a higher sand content than the other sample units.The greater sand content should have caused greater variation in the soil moisture content as well as the EC a values.However, despite the temporal instability (CVi > 30%) shown in the analysis, these sample units presented lower EC a values than the other sample units (Figure 2).For sample units 28 to 50, the CVi for the EC a remained less than 30% in the three soil layers (Figure 3B), demonstrating its temporal stability.Moreti et al. (2007) used the same methodology to assess the space-time behavior of gravimetric and volumetric water storage in an Oxisol cultivated with citrus.They found high correlation coefficients for these attributes over a period of three years of data collection.Cambouris et al. (2006) evaluated the temporal stability of EC a at the soil surface and in the 0 -20 cm soil layer in two different years and obtained significant coefficients of correlation equal to 0.78 and 0.90.These authors emphasized that the strong correlations between the different EC a measurements indicated low temporal variability, suggesting that EC a is a good metric for identifying temporally stable management zones.In a study of the patterns of EC a in coastal saline soil in a province of China, Li et al. (2007), found that regions with a higher salinity exhibited EC a temporal variability (0 < CVi < 30%), whereas regions with low salinity patterns exhibited unstable temporal variability (CVi > 30%).This behavior is different from that observed in this present study.The reason for this difference may be linked to the difference in soil composition.The CVi determination approach has been used to analyze not only the temporal variability of the EC a but also the temporal stability of crop yields (BLACKMORE, 2000); soil attributes such as pH, P, K and Mg (SHI et al., 2002); dry matter yield and nitrogen content in forage grasses (XU et al., 2006); and yield and phosphorus concentration in pastures (SERRANO et al., 2011).

Spatio-temporal variability of the apparent soil electrical conductivity
Visual assessments of the EC a spatiotemporal variability maps revealed a similar pattern of spatial variability on different dates and in different layers of analyzed soil in both Field 1 (Figure 4) and Field 2 (Figure 5).These maps show that the three classes of normalized EC a changed only slightly over time.The location of the normalized EC a values in the 0 -33% class is more stable than in the other two bands.The results were similar to those reported by Farahani and Buchleiter (2004)

Figure 1 .
Figure 1.Thematic maps of apparent soil electrical conductivity (EC a ) measured in the 0 -20 cm soil layer; the respective sampling unit locations are shown for Field 1 (A) and Field 2 (B).

Figure 3 .
Figure 3. Temporal stability of the apparent soil electrical conductivity established based on the CV i values (coefficient of variation for the 20 different dates for the i th sample unit) in Field 1 (A) and Field 2 (B).
in a study of three irrigated sandy soil fields; they verified that classes with low, medium and high values of normalized EC a did not change location over a period of four years.The temporal stability… MEDEIROS, W. N. et al.Biosci.J., Uberlândia, v. 32, n. 1, p. 150-159, Jan./Feb.2016

Figure 4 .
Figure 4. Representation of the spatial pattern of variation of the apparent soil electrical conductivity (EC a ) on the 20 measurement dates for each of the 20 sampling units of Field 1 in the 0 -20 cm (A), 0 -40 cm (B) and 0 -60 cm soil layers (C).

Figure 5 .
Figure 5. Representation of the spatial pattern of variation of the apparent soil electrical conductivity (EC a ) on the 20 measurement dates for each of the 20 sampling units of Field 2 in the 0 -20 cm (A), 0 -40 cm (B) and 0 -60 cm soil layers (C).