Comparison of Soil EC Values from Methods Based on 1:1 and 1:5 Soil to Water Ratios and ECe from Saturated Paste Extract Based Method

The present study investigates the effect of three different methods of obtaining 1:1 and 1:5 soil-over-water mass ratios (soil:water) extracts for soil electrical conductivity (EC) measurements (EC1:1, EC1:5). On the same soil samples, also the electrical conductivity of the saturated paste extract (ECe) was determined and the relationships between ECe and each of the three of EC1:1 and EC1:5 values were examined. The soil samples used were collected from three areas over Greece (Laconia, Argolida and Kos) and had ECe values ranging from 0.611 to 25.9 dS m−1. From the results, it was shown that for soils with ECe < 3 dS m−1 the higher EC values were obtained by the method where the suspension remained at rest for 23 hours and then shaken mechanically for 1 h. On the contrary, no differences were observed among the three methods for soils with ECe > 3 dS m−1. Also, in the case of EC1:5, the optimal times for equilibration were much longer when ECe < 3 dS m−1. Across all soils, the relationships between ECe and each of three methods of obtaining EC1:1 and EC1:5 were strongly linear (0.953 < R2 < 0.991 and 0.63 < RMSE < 1.27 dS m−1). Taking into account the threshold of ECe = 3 dS m−1, different ECe = f(EC1:5) linear relationships were obtained. Although the linear model gave high values of R2 and RMSE for ECe < 3 dS m−1, the quadratic model resulted in better R2 and RMSE values for all methods examined. Correspondingly, in the 1:1 method, two of the three methods used exhibited similar slope values of the linear relationships independent of ECe value (ECe < 3 or ECe > 3 dS m−1), while one method (23 h rest and then shaken mechanically for 1 hour) showed significant differences in the slopes of the linear relationships between the two ranges of ECe.


Introduction
Soil salinity is one of the basic limiting factors in food production especially in arid and semi-arid regions since most crops are sensitive to increased salt concentration in the soil solution [1]. Soil salinization is particularly acute in arid and semi-arid areas with shallow groundwater as well irrigation water of poor quality.
Soil salinity assessment is based on measurement of the electrical conductivity of soil saturated paste extract (EC e ); this has been established as the standard method [2,3]. Saline soils are considered to be the soils where the saturated paste extract has EC e values greater than 4 dS m −1 . However, this method is laborious and time consuming especially in the case of EC e determination for a large number of soil samples. Additionally, the method appears to be more difficult and requires skills and expertise to obtain saturation point for clay soils.
For these reasons, many researchers have suggested easier methods to determine EC in various soils over water mass ratios extracts instead of determining EC e . The most widely used soil over water Table 1. Relationships between soil saturated paste extract electrical conductivity (EC e ) and 1:1 and 1:5 soil to water extract electrical conductivities (EC 1:1 , EC 1:5 ) as proposed by several researchers, as well as the extraction method and the corresponding range of EC e values.
Still now, no comparison has been made among the three abovementioned widely spread EC methods. Also, from international literature, it seems that there is no research work referred on the effect of different methods on the EC 1:1 , although different methods have been used on the EC 1:1 measurement [16,17].
The objectives of present work are: (i) The comparison of EC values derived from the three most commonly used methods of 1:1 and 1:5 extracts; to investigate whether the differences between these methods are maintained across a range of soil EC e and (ii) the investigation of the relationship between EC e and EC values derived from the three methods.

Sample Collection Areas
The soil samples examined were collected from three areas in Greece, and more specifically, from the Prefectures of Lakonia, Argolida and from the island of Kos. Specifically, 50 soil samples were collected from Laconia from irrigated olive groves. The sampling procedure was carried out in September after the irrigation period. In Argolida, 12 samples were collected from various irrigated crops at the end of the irrigation period, while in Kos, 27 samples were collected from a horticultural greenhouse. The depth of soil samples collection was up to 30 cm.

Methods of Determining the Soil Properties
After sampling, the samples were transferred to the laboratory for air-drying and sieving through a 2 mm sieve and the soil texture, pH and calcium carbonate were determined. Soil texture was determined by means of the Bouyoucos hydrometer method [23], pH values were measured using standard glass/calomel electrodes in 1:2.5 w/v soil-water suspension [24]; CaCO 3 equivalent percentage was estimated by measuring the eluted CO 2 following the addition of HCl (calcimeter Bernard method).

Methods of Various Soil Extraction and Measurements
2.3.1. EC e Method 350 g of soil was used to prepare the soil saturated paste and then the paste was allowed to stand for 24 h (USDA, 1954). Subsequently, the vacuum extracts were collected and EC e was measured by a conductivity meter (WTW, Cond 315i). For the saturation percentage (SP) determination, a subsample of each paste was oven dried at 105 • C for 24 h.

EC 1:5 Method
For the 1:5 suspension, 50 g of soil and 250 mL of distilled water were used. Three alternative methods were applied: the method of Loveday [18], the NRCS [15] and the USDA [2].
In the Loveday method, the suspension was shaken by a mechanical shaker for exactly one hour and then kept at rest for 20 min. After the rest time, the extract was obtained, and the EC was determined. For the NRCS method, the suspension remains at rest in complete shade for 23 h and then shaken mechanically for one hour. After the shaking, the extract was obtained, and the EC was determined. Finally, in the USDA method the suspension was shaken by hand, 4 times, every half hour for 30 s. After, the extract was obtained, and the EC was determined. The method of vacuum filtration in all the three methods is the same and common, followed by the measurement of EC with a conductivity meter. All the methods and EC readings were conducted at 25 • C.
In two soil samples, one from Laconia (sample L) and one from Argolida (sample A) with EC e values of 0.793 and 13.78 dS m −1 , respectively, the EC 1:5 values were measured after the suspensions were agitated with mechanical shaker for times 1, 2, 3, 4, 6, 24 and 48 h. After each agitation time the extraction was obtained, and the EC was determined. This process can better evaluate the role of shaking time on the EC 1:5 values for the two very different EC e values.

EC 1:1 Method
In the 1:1 method, the three above mentioned methods (Loveday, NRCS and USDA) were also applied as described in the 1:5 method. For each of the above methods, 50 g of soil was weighed and then each procedure was performed in the same way as above.

Statistical Analysis
For the relationships EC e = f(EC 1:1 ) and EC e = f(EC 1:5 ), a least-squared linear regression was applied and the coefficient of determination R 2 was evaluated. The R 2 coefficient is used to assessing the correlation between two independent methods. Also, the values of root mean square errors (RMSE) were determined. Analysis of variance (ANOVA) was applied to test the significant difference among the applied EC 1:5 or EC 1:1 methods using SPSS Statistical Software v. 17.0 (SPSS Inc., Chicago, IL, USA); the means of each method were compared using t-test at a probability level P = 0.05.

Soil Properties
Samples from Laconia and Argolida are characterized as clay-clay loam soils and from Kos as sandy clay soils. All soil samples presented negligible gypsum content. As regards to CaCO 3 , samples from Laconia presented a content lower than 2.5%, from Argolida 5-8% and from Kos 8.5-11%. The pH values ranged from 7.69 to 8.06 for soil samples from Laconia and from 7.5 to 7.7 for soil samples from Argolida and Kos.
Additionally, the soil texture analyses of the two soil samples examined separately resulted as follows: (i) soil sample L-clay soil (23.5% sand, 16% silt, 60.5% clay) and (ii) soil sample A-clay loam/loam soil (39% sand, 32% silt, 29% clay). The CaCO 3 content was 0.2% and 7.66% and pH values were 7.75 and 7 for sample L and A, respectively.

Estimation of Soil Salinity
The EC e values ranged from 0.611 to 25.9 dS m −1 . It should also be noted that the EC e variation range of the soil samples from Laconia is much lower than that of the other two regions (Argolida and Kos). Specifically, EC e values of the samples from Laconia ranged from 0.611 to 1.664 dS m −1 , while in the other two regions they ranged from 2.32 to 25.9 dS m −1 . From the measured EC e values, it appears that a relatively wide range in salinity levels was obtained for both comparing the different EC 1:5 and EC 1:1 methods, as well as evaluating the relationship between the EC e and each of EC 1:5 or EC 1:1 methods.
As regards to SP all soil samples examined (with exception of the two separated samples) have values greater than 43%, percentage which indicates that the soils are classified in fine textured soils [20]. More specifically, SP values ranged from 50.5% to 72.5% for soils from Laconia, 52-70% for soils from Argolida and 43-53% for soils from Kos.

Comparison of 1:1 and 1:5 Soil to Water Extract Electrical Conductivity Methods
In Table 2 the slope of the linear relationship (y = ax) between 1:5 soil to water extract electrical conductivity methods for EC e < 3 dS m −1 and EC e > 3 dS m −1 and R 2 are presented. Similarly, the slope and R 2 of the linear relationship between 1:1 soil to water extract electrical conductivity methods for EC e < 3 dS m −1 and EC e >3 dS m −1 are presented in Table 3. Table 3. Slopes of the linear equations describing the relation between 1:1 soil to water extract electrical conductivity methods for EC e < 3 dS m −1 and EC e > 3 dS m −1 and coefficient of determination R 2 . From the results presented in Tables 2 and 3, it is obvious that each of the three methods examined resulted in different values of both EC 1:1 and EC 1:5 when EC e < 3 dS m −1 . Analysis of variance (ANOVA) showed that the three methods are significantly different at a probability level P = 0.05. Furthermore, the t-test analysis (P = 0.05) showed that the NRCS and Loveday methods as well as the USDA and Loveday methods resulted in significantly different EC 1:5 values, while EC 1:5 values between NRCS and USDA were not significantly different. The mean value with standard deviation for NRCS, USDA and Loveday methods were 0.177 ± 0.029, 0.169 ± 0.029 and 0.151 ± 0.027 dS m −1 , respectively. In the case of 1:1 ratio, the EC values between NRCS and USDA as well as NRCS and Loveday methods were also significantly different (P = 0.05). The mean value with standard deviation for NRCS, USDA and Loveday methods were 0.5 ± 0.070, 0.43 ± 0.100 and 0.423 ± 0.086 dS m −1 , respectively.
The NRCS method resulted in greater EC values compared to the other two methods for both 1:1 and 1:5 ratios, whereas the Loveday method resulted in lower EC values. From these results, it appears that at low values of EC e (EC e < 3 dS m −1 ) the rest time seems to play an important role since the difference between the NRCS and the Loveday method is only in the duration of rest time. As regards to the NRCS and USDA methods, the slope of the linear regression between the NRCS and USDA at 1:5 ratio is 1.047, while at 1:1 is 1.161.
The Therefore, it appears that the agitation time plays a dominant role to obtain equilibrium since the difference between the NRCS method (EC 1:5 = 0.158 dS m −1 ) and the method with 24 h shaking (EC 1:5 = 0.218 dS m −1 ) is in the shaking time. These results are similar to those of He et al. [6] in terms of the long shaking time required to equilibration but differ in the fact that in our experiments did not show differences in EC values obtained by shaking of at least up to 6 h. He et al. [6] explained that the higher values of EC obtained by the long shaking time method compared to other methods may be due to the fact that the mechanical shaking destroys micro-aggregates, as well as increase dissolution of salts because the dynamic concentration gradient between solid and liquid phases. Also, Vanderheynst et al. [12] found that differences occur for shaking time greater than a threshold value of 3 h.
In the case of soils with EC e > 3 dS m −1 there is no significant differences between agitation methods since all methods gave almost the same results and the slope of the linear relationship is almost 1 (Tables 2 and 3). In addition, it is noted that the R 2 values for soils with EC e > 3 dS m −1 are higher for all methods examined, in both 1:5 and 1:1 ratios, compared to R 2 values for EC e < 3 dS m −1 (Tables 2 and 3).
The EC 1:5 values of the soil sample A (with EC e = 13.8 dS m −1 > 3 dS m −1 ) obtained by mechanical shaking for 1, 2, 3, 4, 6, 24 and 48 h ranged from 1.683 to 1.751 dS m −1 . It is obvious that for soils with EC e > 3 dS m −1 the shaking times required to obtain equilibration are significantly lower compared to soils with EC e < 3 dS m −1 The different behavior depending on the EC e value shows that the solid and liquid phases is far from considered a simple system where the only process carried out is dissolution and that the concentration of ions is inversely proportional to dilution. Such situations may exist only in sandy or sandy loam soils in semi-arid areas with high salinity [25]. However, the soils are characterized by a cation exchange capacity value depending on the type and quantity of clay, the presence of slightly soluble minerals but also ion exchanges between solid and liquid phase. In the present experimental work, the existence of a relatively high clay percentage combined with the existence of slightly soluble minerals may be led to different EC values among various methods, especially when EC e < 3 dS m −1 . This phenomenon may be even more pronounced in the case of clay soils where there are high content of slightly soluble minerals but less pronounced in the coarse-textured soils without slightly soluble minerals.

Relationship between EC e and 1:5 Soil to Water Extract Electrical Conductivity Methods
In Table 4, the linear relationships between EC e and EC 1:5 , for all soil samples, determined by the three different methods are presented. Analysis of the results showed that each 1:5 soil to water extract electrical conductivity method is strongly related with EC e since R 2 values are high (0.953 < R 2 < 0.972) and RMSE are low (1.02 dS m −1 < RMSE < 1.27 dS m −1 ). It also appears that the linear equations showed small differences regardless of the EC 1:5 methods for all soils examined. These data confirm the existence of a strong linear relationship when the range of EC e is relatively great (Table 1). As shown in Table 4, the relationship EC e = fEC 1:5 using the USDA method is similar to the corresponding one reported by Kargas et al. [7], (Table 1) for Greek soils since both the two equations have almost the same slope (6.61 and 6.53, respectively).
However, analysis of the results for soils with EC e < 3 dS m −1 showed that a percentage of 70% of experimental EC e values were lower than those calculated by the equations presented in Table 4. For this reason, the data were separated into two ranges based on the threshold value EC e = 3 dS m −1 to evaluate whether the relationship EC e = fEC 1:5 is described by different equations as reported by other researchers [26,27].
The slopes of linear equation describing the relation between EC e and EC 1:5 determined by three different methods, as well as the R 2 and RMSE for all soil examined for EC e < 3 dS m −1 and EC e > 3 dS m −1 , are presented in Table 5. Table 5. Regression equations describing the relation between saturated paste extracts EC e and EC 1:5 determined by three different methods with the coefficients of determination (R 2 ) and root mean square errors (RMSE) for all soil examined for EC e < 3 dS m −1 and EC e > 3 dS m −1 . As shown in Table 5, for soils with EC e < 3 dS m −1 , the slope of the linear equation between EC e and EC 1:5 has different value depending on EC 1:5 determination method used with the smallest and the highest values obtained by the NRCS and Loveday method. Also, the values of the slopes of linear relationships, for both EC e < 3 dS m −1 and EC e > 3 dS m −1 , differ significantly from each other since in the case of EC e < 3 dS m −1 these values ranged from 4.68 to 5.46, while they ranged from 6.60 to 6.71 in the case of EC e > 3 dS m −1 . In addition, for EC e < 3 dS m −1 R 2 values are lower (0.537 < R 2 < 0.718) Water 2020, 12, 1010 8 of 12 than those ones (0.917 < R 2 < 0.942) observed for EC e > 3 dS m −1 indicating a strong linear relation between EC e and each EC 1:5 determination method.
Comparison between the same methods for both EC e < 3 dS m −1 and EC e > 3 dS m −1 showed a difference between slopes ranging from 18.5% to 28.9%. Thus, in order to compare various equations describing the relationship between EC e and EC 1:5 , both the agitation method of EC 1:5 determination and the range of EC e for which the equation has been proposed should be taken into account. Specifically, as shown in Table 5 and Figure 1, the relationship between EC e and EC 1:5 determined by the NRCS method has a slope of 4.68 for EC e < 3 dS m −1 and 6.60 for EC e > 3 dS m −1 . The differences among the methods may be even greater if the soil contains gypsum or larger amounts of calcite than those observed in the soil samples examined. Table 5, for soils with ECe < 3 dS m −1 , the slope of the linear equation between ECe and EC1:5 has different value depending on EC1:5 determination method used with the smallest and the highest values obtained by the NRCS and Loveday method. Also, the values of the slopes of linear relationships, for both ECe < 3 dS m −1 and ECe > 3 dS m −1 , differ significantly from each other since in the case of ECe < 3 dS m −1 these values ranged from 4.68 to 5.46, while they ranged from 6.60 to 6.71 in the case of ECe > 3 dS m −1 . In addition, for ECe < 3 dS m −1 R 2 values are lower (0.537 < R 2 < 0.718) than those ones (0.917 < R 2 < 0.942) observed for ECe > 3 dS m −1 indicating a strong linear relation between ECe and each EC1:5 determination method.

As shown in
Comparison between the same methods for both ECe < 3 dS m −1 and ECe > 3 dS m −1 showed a difference between slopes ranging from 18.5% to 28.9%. Thus, in order to compare various equations describing the relationship between ECe and EC1:5, both the agitation method of EC1:5 determination and the range of ECe for which the equation has been proposed should be taken into account. Specifically, as shown in Table 5 and Figure 1, the relationship between ECe and EC1:5 determined by the NRCS method has a slope of 4.68 for ECe < 3 dS m −1 and 6.60 for ECe > 3 dS m −1 . The differences among the methods may be even greater if the soil contains gypsum or larger amounts of calcite than those observed in the soil samples examined. Similar results regarding to the effect of agitation method, the range of ECe and the gypsum content on equation describing the relationship between ECe and EC1:5 have been presented by other researchers [3,26,27].
He et al. [27] proposed a quadratic equation as a more appropriate equation to describe the relationship between ECe and EC1:5 when ECe values are lower than 4 dS m −1 . The fitting of a quadratic equation to the data of this study for ECe < 3 dS m −1 gave R 2 values of 0.74, 0.57 and 0.66 and RMSE values 0.096 (NRCS), 0.124 (USDA) and 0.115 dS m −1 (Loveday method), respectively. A comparison between these RMSE values and those of the linear relationships presented in Table 5, showed a significant improvement only in the case of the NRCS method. It should be noted that there is a significant difference in RMSE values presented in Table 4 compared to RMSE values whether we use the linear equation or quadratic equation to ECe estimation for ECe < 3 dS m −1 . Similar results regarding to the effect of agitation method, the range of EC e and the gypsum content on equation describing the relationship between EC e and EC 1:5 have been presented by other researchers [3,26,27].

Relationship between ECe and 1:1 Soil to Water Extract Electrical Conductivity Methods
He et al. [27] proposed a quadratic equation as a more appropriate equation to describe the relationship between EC e and EC 1:5 when EC e values are lower than 4 dS m −1 . The fitting of a quadratic equation to the data of this study for EC e < 3 dS m −1 gave R 2 values of 0.74, 0.57 and 0.66 and RMSE values 0.096 (NRCS), 0.124 (USDA) and 0.115 dS m −1 (Loveday method), respectively. A comparison between these RMSE values and those of the linear relationships presented in Table 5, showed a significant improvement only in the case of the NRCS method. It should be noted that there is a significant difference in RMSE values presented in Table 4 compared to RMSE values whether we use the linear equation or quadratic equation to EC e estimation for EC e < 3 dS m −1 . Table 6 shows the relationship between EC e and the three methods of determining EC 1:1 for all soil samples examined. The results showed that the relationship is strongly linear in all methods examined (R 2 > 0.986) and RMSE values are low (0.63 < RMSE < 0.74 dS m −1 ). The values of both R 2 and RMSE indicate that this linear relationship reliably estimates the EC e . However, EC e = fEC 1:1 linear relationships have different f coefficient for each method. In Table 7, regression equations describing the relation between EC e and EC 1:1 determined by three different methods are presented taking into consideration the threshold of EC e value 3 dS m −1 . The results showed that the same trends were observed for R 2 and RMSE values as in the case of the results of 1:5 ratio presented in Table 5. As regards to differences observed in the slope of linear relationships between the two areas of EC e values, a notable difference was observed in the NRCS method since it resulted to a slope 1.65 for EC e < 3 dS m −1 and 2.08 for EC e > 3 dS m −1 . Furthermore, the quadratic equation for the NRCS method, for EC e < 3 dS m −1 , resulted almost to the same RMSE values (0.099 dS m −1 ) with those of linear equation. Therefore, for this method with EC e <3 dS m −1 the simple linear equation gave quite reliable results to EC e estimation. The other two methods showed similar slope values regardless of the EC e value. In particular, the EC e -USDA relationship had almost the same slope value regardless of the EC e . Table 7. Regression equations describing the relation between saturated paste extracts EC e and EC 1:1 determined by three different methods with the coefficients of determination (R 2 ) and root mean square errors (RMSE) for all soil examined for EC e < 3 dS m −1 and EC e > 3 dS m −1 . The relationships between EC e and EC 1:1 determined by the NRCS method taking into consideration the threshold of EC e value 3 dS m −1 are also presented in Figure 2. The relationships between ECe and EC1:1 determined by the NRCS method taking into consideration the threshold of ECe value 3 dS m −1 are also presented in Figure 2.

Conclusions
The EC1:5 was affected by both agitation method and time, especially for ECe values lower than 3 dS m −1 . Generally, the NRCS method resulted in the highest EC values compared to the other two methods examined. The differences among agitation methods are essentially eliminated for ECe values greater than 3 dS m −1 . For soil having ECe values lower than 3 dS m −1 , equilibration time was very greater than the soils having ECe values above 3 dS m −1 . The most appropriate equation for ECe estimation using EC1:5 values for soils having ECe < 3 dS m −1 is a quadratic equation-especially in the case of the NRCS method-while for soils having ECe > 3 dS m −1 is the linear equation. However, if soils have a wide range of salinization levels, the linear model are recommended.
The present study shows that the shaking method and the equilibration time are additional contributing factors to the observed differences of the proposed equations for the ECe estimation by EC1:5. Therefore, in order to select each time, the appropriate method and equilibration time for measuring EC1:5, during laboratory studies, the ECe value of some samples, as well as the soil characteristics (e.g., gypsum and calcium carbonate content) should be examined in advance.
The EC1:1 was affected by ECe values only in the case of the NRCS method where the estimation of the ECe can be conducted by simple but different linear relationships whose slopes depend on ECe values. In the other two methods, the linear relationship ECe = f (EC1:1) was not affected by ECe values.
Overall, it is necessary to describe in detail the method of preparation and extraction for determining EC1:1 or EC1:5 and the range of ECe in order to properly evaluate and compare the proposed equations of ECe = f(EC1:5). Additionally, the study of soils with different characteristics than those of the group of soils examined in this work is needed.

Conclusions
The EC 1:5 was affected by both agitation method and time, especially for EC e values lower than 3 dS m −1 . Generally, the NRCS method resulted in the highest EC values compared to the other two methods examined. The differences among agitation methods are essentially eliminated for EC e values greater than 3 dS m −1 . For soil having EC e values lower than 3 dS m −1 , equilibration time was very greater than the soils having EC e values above 3 dS m −1 . The most appropriate equation for EC e estimation using EC 1:5 values for soils having EC e < 3 dS m −1 is a quadratic equation-especially in the case of the NRCS method-while for soils having EC e > 3 dS m −1 is the linear equation. However, if soils have a wide range of salinization levels, the linear model are recommended.
The present study shows that the shaking method and the equilibration time are additional contributing factors to the observed differences of the proposed equations for the EC e estimation by EC 1:5 . Therefore, in order to select each time, the appropriate method and equilibration time for measuring EC 1:5 , during laboratory studies, the EC e value of some samples, as well as the soil characteristics (e.g., gypsum and calcium carbonate content) should be examined in advance.
The EC 1:1 was affected by EC e values only in the case of the NRCS method where the estimation of the EC e can be conducted by simple but different linear relationships whose slopes depend on EC e values. In the other two methods, the linear relationship EC e = f(EC 1:1 ) was not affected by EC e values.
Overall, it is necessary to describe in detail the method of preparation and extraction for determining EC 1:1 or EC 1:5 and the range of EC e in order to properly evaluate and compare the proposed equations of EC e = f(EC 1:5 ). Additionally, the study of soils with different characteristics than those of the group of soils examined in this work is needed.