International Journal of Agriculture and Crop Sciences.

Available online at www.ijagcs.com

IJACS/2012/4-4/157-167
ISSN 2227-670X 2012 IJACS Journal


Comparison of interpolation methods for estimating
pH and EC in agricultural fields of Golestan province
(north of Iran)

Hossein Kazemi Poshtmasari
1
, Zeynolabedin Tahmasebi Sarvestani
2
*, Behnam Kamkar
3

and Shaban Shataei
4
, Sohrab Sadeghi
5

1. PhD Student of Tabiat Modares University, Tehran, Iran.
2. Scientific Member of Tabiat Modares University, Tehran, Iran. (Corresponding author)
3. Scientific Member of Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran.
4. Gorgan University of Agricultural Sciences and Natural Resources, PO Box 49138-15739 ,Gorgan, Iran.
5. Scientific Member of Gorgan University of Agricultural Sciences and Natural Resources, Gorgan Iran.

*Corresponding author email: tahmaseb@modares.ac.ir

ABSTRACT: In this research, the spatial distribution of Electrical Conductivity (EC) and pH on
agricultural fields of Golestan province (north of Iran) were evaluated using some interpolation methods
such as Kriging with different models, Inverse Distance Weighted (IDW) and Radial Basis Function
(RBF). For this purpose, 505 soil samples were provided from fields during 2010. EC and pH rates were
measured for each sample. The performance of methods was evaluated using by Mean Absolute Error
(MAE), Mean Bias Error (MBE) and Root Mean Square Error (RMSE). The results showed that
geostatistical methods had a higher accuracy compared to IDW and RBF. Kriging (Spherical model)
was the best method to estimate pH, because it had the highest precision and lowest error. Exponential
model-based kriging had the highest precision for estimation of soil EC in this area. Also, Radial Basis
Function (TPS) model was the most unsuitable method for estimation of these soil properties.
Semivariograms analysis indicated that pH and EC were best fitted to Spherical model

Key Words: Agricultural fields, electrical conductivity, IDW, Kriging, RBF

INTRODUCTION

In a division of Caspian region of Iran, Golestan province was established with Gorgan city as capital in
1997. It is a transition area from arid and semi-arid to temperate-humid. Most of the province is under the influence
of the Caspian Sea. These characteristics of Golestan province have a great influence in the variety of cultivated
plants. Golestan is very famous in view of agricultural productions in Iran. In this province, about 600,000 hectares
are cultivated and wheat, barley, soybean, canola, cotton and rice are the main crops. Uncultivated areas are
covered with pastures and forests. Wheat and barely are generally cultivated in a central band with a width of up to
20 Km where it extends from low altitude coastal areas in west to high mountains in north and northeast. Recently,
this province is faced with a lot of salinity related problems in north and northeast and agricultural area became the
worst affected.
Soil salinity assessment is one of the important components in agriculture management and water
allocation strategies. There are three significant impacts on soil and plants when salined water is used for irrigation.
The first impact is that soil particles are dispersed and because of this aggregation is destroyed. The second one is
that ability of plants to uptake water will reduce by increasing osmotic potential. The third impact is that the ionic
balance of the soil solution will affecte by reducing nutrient absorption (Brady and Wells 2001; Gawel 2009;
Pisinaras et al., 2010). Mapping soil salinity, demands a considerable effort in taking soil samples and measuring
Electrical Conductivity (EC) (Utset et al., 1998). The relationship between EC and soil salinity is complicated by
other factors influencing measured conductivity in the field, such as soil texture, water content, and bulk density
(Banton et al., 1997). Knowledge of soil salinity and its variability is essential in order to evaluate the extent of salt
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

138

build-up and to recommend appropriate management practices to increase crop yield. The soil salinization
processes may operate at a local or regional scale (Castrignano et al., 2008).
Under arid or semi-arid conditions and in regions of poor natural drainage, there is a real hazard of salt
accumulation in soils. The processes by which soluble salts cause salinity and sodicity in soils include the
application of waters containing salts, weathering of primary and secondary minerals in soils, organic matter decay
and watertable instability. The importance of each of these causes depends on soil type, climate and agricultural
managements. Accumulation of dispersive cautions such as Na in soil solution and the exchange phase (K, Mg,
Ca) affect the physical properties of soil, such as structural stability, hydraulic conductivity, infiltration rate and
erosivity (Juan et al., 2010). Corwin and Lesch (2010) provided the protocols for conducting a field-scale EC survey
and applied these protocols to a soil quality assessment in central Californias San Joaquin Valley. The protocols
comprise eight general steps:(a) site description and EC survey design; (b) EC data collection with mobile GPS-
based equipment;(c) soil sampling design; (d) soil core sampling; (e) laboratory analysis; (f) calibration of EC (g)
spatial statistical analysis; (h) GIS Database development and graphic display. The developed protocols provide
the guidelines to assure reliability, consistency and compatibility of EC survey measurements and their
interpretation (Juan et al., 2010).
In recent years, geostatistics has been proven to effectively assess the variability of soil properties
(Webster, 1994; Zhang et al., 1998; Zhang et al., 2000; Webster and Oliver, 2001; Corwin et al., 2003; Mueller et
al., 2003; Sun et al., 2009). Geostatistics provides a set of statistical tools for analyzing spatial variability and spatial
interpolation. These techniques produce not only prediction surfaces but also error or uncertainty surfaces.
A semivariogram is used to describe the structure of spatial variability. The semivariogram plays a central
role in the analysis of geostatistical data using the kriging method. It takes into account the spatial autocorrelation in
data to create mathematical models of spatial correlation structures commonly expressed by variograms
(Gundogdu and Guney, 2007; Uyan and Cay, 2010). The interpolation method is recognized as a critical factor for
interpolation accuracy. Interpolation methods can be classified into three categories: statistics (regression tree and
multiple regression), geostatistics (ordinary kriging and universal kriging), and hybrid (co-kriging and regression
kriging). Therefore, less information can be interpolated for guiding the selection of interpolation method for wide-
ranging soil properties within the same field or for the same soil properties across contrasting landscapes (Zhu and
Lin, 2010).
Radial Basis Functions (RBF) networks are a special type of neural networks (RBFNN) which employ both
local and global functions to provide spatial estimates. They are easily adaptive to the monitoring networks by
means of adjusting kernels positions. Geostatistical tools are used to control the quality of RBFNN mapping (Lin
and Chen 2004; Webb and Shannon, 1998).
The accuracy of interpolation methods for spatially predicting soil and water properties has been analyzed
in several studies. Pozdnyakova and Zhang (1999) used kriging and cokriging methods to estimate sodium
adsorption ratio (SAR) in a 3375 ha agricultural field. In cokriging, measured data of electrical conductivity (EC)
were incorporated to improve the estimation of SAR. The results suggested that sampling cost can be dramatically
reduced and estimation can be significantly improved using cokriging. Taghizadeh Mehrjardi et al., (2008) in
Ardakan-Yazd plain of Iran applied the Inverse Distance Weighted (IDW), kriging and cokriging methods for
predicting spatial distribution of some groundwater characteristics such as: EC, SAR, Cl
-
and SO
4
2-
. Results showed
that kriging and cokriging methods are superior to IDW. The cokriging has higher accuracy than other methods for
estimating spatial distribution of groundwater quality. Mashayekhi et al., (2007) investigated regional variation of
salinity and sodicity in southern Lenjan-Isfahan (Iran) with geostatistical methods. Soil samples were obtained from
0-0.2, 0.2-0.5 and 0.5-1.0 m depths. EC and SAR maps in all depths were prepared by kriging and cokriging
methods. Maroofi et al., (2009) indicated that local polynomial method and Inverse Distance were the best methods
to estimate EC and pH, respectively, in stream drained water in Hamedan-Bahar plain, west of Iran. In this study,
RBF method was also found as the most unsuitable method.
Istock and Cooper, (1998) used kriging method to estimate heavy metals and found that the used method is
the best estimator for spatial prediction of metals. In another research, spatial distribution maps were constructed
for EC and pH of soil extracts using ordinary kriging interpolation in the agricultural lands of Rhodope District,
northeastern Greece (Pisinaras et al. 2010). Yu et al., (2008) analyzed the distribution of surface soil pH by
combining classical statistics method with geostatistic method under three irrigation methods (furrow, drip and
subsurface irrigation) in greenhouse. The results indicated that supplying method and quantity of water by irrigation
can affect the spatial variability of soil pH. The results of study in Wuhu county in China showed that local
polynomial method, ordinary kriging, simple kriging and disjunctive kriging had higher spatial prediction accuracy
than the other interpolation methods. The spatial prediction accuracy would increase with increasing sampling from
soil Olsen-P number (Sun et al., 2009). The present study was therefore carried out with the objective of evaluating
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

139

accuracy of different interpolation methods, kriging, RBF and IDW, for analyzing the spatial pattern and mapping of
EC and pH in agricultural fields at province scale, in the Golestan province, north of Iran.

MATERIALS AND METHODS
Study Area
This research was conducted in northern part of Iran, in Golestan province, during the 2010. This province,
with an area of 20,033 km
2
, is located in the southeast of Caspian sea between 36 44 and 38 5 north latitude
and 53 51 and 56 14east longitude (Figure 1). It covers about 1.3% of the total area of the country. It has 270
km
2
international border with Turkmenistan at the north. Also, it is situated in the vicinity of Semnan, Mazandaran
and Northern Khorasan provinces on the south, west and east, respectively. The eastern extension of Alborz
Mountains Range surrounds the coastal plains of the Caspian sea as a high and long wall, thus all over the
province the land slope decreases from the southern and eastern mountains towards sea, with altitude ranging
between -25 to 3000 m. The mountains of Shahkuh, Siahmarzkuh, Chahbid and the mountain chain of Kurkhud
keep its southern and eastern parts with their limits. The some important townships of this province are Gonbad-e-
Kavoos, Minoodasht and Gorgan with an area of 6856.8 km, 6485.6 km and 2848 km, respectively. The climate
of the province is under the influence of Alborz Mountains, Caspian Sea, the southern wildernesses of
Turkmenistan, and forests. According to De-Martonne advanced climate classification system, the province
contains five different climates: Mediterranean in center, arid-desert in north, semi-arid in coast, center and
northeast, humid in sub-south, and semi humid in south.

Data Sampling and Analysis
Soil samples were randomly taken from 505 locations in May 2008. Sampling points are showed in Figure
2. Samples were taken at depths of 030 cm and air-dried to remove stones and coarse crop residues. Three
samples from each sampling site were collected from the surface soil layer and samples were thoroughly mixed
and one sample ground to pass a sieve, then stored in plastic bags prior to analysis. The EC values of sample were
determined by EC meter and pH was measured using a pH meter with a glass electrode.

Geostatistical Analysis
In general, geostatistical methods were used to estimate and map in this agricultural areas. It is based on
the theory of a regionalized variable which is distributed in space (with spatial coordinates) and shows spatial auto
correlation such that samples close together in space are more alike than those that are further apart. Geostatistics
uses the variogram technique (or semivariogram) to measure the spatial variability of a regionalized variable, and
provides the input parameters for the spatial interpolation of kriging (Krige, 1951; Goovaerts, 1999; Webster and
Oliver, 2001).
The semivariogram (Variogram) was used in this study to analyze discrete soil samples. Semivariograms
are a key tool in regionalized variables theory and are formed by three constituents: sill, range and nugget with
increasing lag between samples; semivariance is increased to a maximal asymptotic value (sill). With this lag,
semivariance is approached the observation variance. This lag is called range beyond which variables are
independent with no correlations. Nugget occurs when semivariogram is not started exactly at intersection of
coordinates generally due to laboratory test errors, a sharp variation of soil properties or when sampling distance is
greater than range. Initial slope intensity in semivariogram exhibits variability as a function of distance and
reduction of correlation between samples (Mashayekhi et al., 2007).
Semivariogram is computed as half the average squared difference between the components of data pairs
(Goovaerts, 1999, Webster and Oliver, 2000): The function is expressed as:
(1)
=
Where N(h) is the total number of data pairs separated by a distance; h; Z represents the measured value for soil
property; and x is the position of soil samples.
Before the geostatistical estimation, a semivariogram is calculated for classes of distance between sample
pairs. Several standard models are available to fit the experimental semivariogram, e.g., spherical, exponential,
Gaussian, linear and power models (Shi et al., 2007). In this study, the Exponential, Spherical and Gaussian
models were selected.



Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

160

Interpolation methods
In this study, spatial patterns pH and EC were determined using the geostatistical and interpolation
methods such as, Kriging, Inverse Distance Weighted (IDW), and Radial Basis Functions (RBF).

Kriging
Among the geostatistical techniques, kriging is an important tool in geostatistics. Kriging is a linear
interpolation procedure that provides a best linear unbiased estimation for quantities which vary in space. Kriging
estimates are calculated as weighted sums of the adjacent sampled concentrations. That is, if data appear to be
highly continuous in space, the points closer to those estimated receive higher weights than those farther away
(Cressie, 1990). Kriging (Krige, 1951) is regarded as an optimal method of spatial prediction. It is a theoretical
weighted moving average:
(2)
=
Where is the value to be estimated at the location of x
0
, z(x
i
) is the known value at the sampling site
x
i
and n is the number of sites within the search neighbourhood used for the estimation. The number n is based on
the size of the moving window and is defined by the user. Kriging is different from other methods (such as IDW),
because the weight is no longer arbitrary. The weights depend on the parameters of the semivariogram model and
the sampling configuration and are decided under the conditions of unbiasedness and minimized estimation
variance (Deutsch and Journel, 1998; Zhangand McGrath, 2004; Robinson and Metternicht, 2006).

IDW
In IDW method, a weight is attributed to the point to be measured. The amount of this weight is dependent
upon the distance of the point to another unknown point. These weights are controlled on the bases of power of
ten. Often with increase of power the effect of the points that are farther diminishes. Lesser power distributes the
weights more uniformly between neighboring points. We should keep in mind that in this method the distance
between the points count, so the points of equal distance have equal weights (Burrough and McDonnell, 1998;
Istokand and Cooper, 1998). In this method the weight factor is calculated with the use of the following formula:
(3)



Where
i
,D
i
, i and are the weight of point, the distance between point, the unknown point and the
power ten of weight, respectively (Taghizadeh Mehrjardi et al., 2008).

RBF
Radial Basis Functions (RBF) methods are a series of exact interpolation techniques; that is, the surface
must go through each measured sample value. Each basis function has a different shape and results in a slightly
different interpolation surface. RBF showed that RBFNN gives the same results as a standard geostatistical
technique: prediction mapping of the regression function and mapping of the predictions variance. Radial basis
function neural networks express the regression function in the form:
(4)
Z=
j
;
1,m

j
h
j
+
0
+

Where is a zero-mean noise, and h radial basis functions.

Radial basis functions are a class of functions of which the key feature is that the distance from a center
determines their response. Most commonly used RBF is a Gaussian radial basis function. Gaussian model
provides an example of local RBF that gives significant response only in a neighbourhood of its center (Lin and
Chen 2004; Webb and Shannon 1998; Haykin, 1994; Bishop, 1995).

Comparison among the Different Methods
To evaluate interpolation methods, statistical indices of MAE (Mean Absolute Error), MBE (Mean Bias
Error) and RMSE (Root Mean Square Error) were used. The MAE is an indicator of errors in the results and MBE
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

161

indicates the bias of the results obtained through the applied method. When MAE and MBE are 0.00 or near to
naught, the applied method simulates the fact well. However, as far as its amount is farer than 0.00, it implies to
less precise and more bias. Finally, we use the RMSE to evaluate model performances in cross-validation mode.
The smallest RMSE indicate the most accurate predictions. How the parameters MAE and MBE and RMSE are
calculated, has been indicated as Eq 5-7:
MBE= (5)

MAE= (6)
(7)

Where: Z(x
i
) is observed value at point x
i
, Z*(x
i
) is predicted value at point x
i
, N is the number of samples.

The validation and the sufficiency of the developed model variogram can be tested via a technique
called cross validation. Cross validation estimation is obtained by leaving one sample out and using the
remaining data. This test allows to assess the goodness of fitting of the variogram model, the appropriateness of
neighbourhood and type of kriging used. The interpolation values are compared to the real values and then the
least square error models are selected for regional estimation (Leuangthong et al., 2004; Uyan and Cay, 2010).

Data Analysis
Statistical results indicated that the soil pH was normally distributed, EC was fitted for normal logarithmic
distribution (Figure 3). Data sets were analyzed with different software packages. Maps were produced with GIS
software ArcGIS and its extension of Spatial Analyst. The geostatistical analysis and the probability calculations
were carried out with GS
+
and geostatistic extension of ArcMap.

RESULTS AND DISCUSSION

Table 1 lists the summary statistics of the raw and log-transformed EC and pH data, including mean,
maximum, minimum, standard deviation, skewness, and kurtosis. Kurtosis (9.31) and skewness (2.92) values for
EC were high and the raw data sets were logarithmically transformed before performing geostatistical analysis.
Logarithmic transformation resulted in reduced skewness (0.96) and kurtosis (-0.27) values and the transformed
data sets passed the lognormal tests. The high value of skewness is demonsrator showing insufficient number of
samples and unsuitable distribution. Histograms of raw and log-transformed values of EC showed in Figure 3.

Table 1. The statistical values of soil properties






Figure 4 presents the semivariogram and fitted model for pH. The attributes of the semivariograms for soil
pH are summarized in Table 2. Preliminary calculations of variograms in different directions showed that all
semivariogram were isotropic. Semivariograms analysis indicated that pH was best fitted to Spherical model with
nugget, sill, and nugget/sill equal to 0.021, 0.047, and 0.562, respectively. In this research, nugget/sill ratio (56.2%)
indicated moderate spatial dependence at the large scale of the Golestan province.
The results of geostatistical analyses of pH have been presented in Table 3. The results showed that
kriging (Spherical model) was the best method to estimate pH, because it had the highest precision and lowest
error for estimation of these elements. The performance of three models (Spherical, Exponential and Gaussian)
have been compared. According to the cross-validation parameters, generally all three models performed fairly well
Soil properties Mean Min Max Standard
Deviation
Skewness Kurtosis
pH 7.75 7.0 8.50 0.218 -0.89 1.97
EC(Log transformatied) 0.366 -1.90 51.0 1.310 0.96 -0.27
EC(untransformatied) 4.085 0.150 3.95 7.429 2.92 9.31
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

162

but Spherical was the best model. Radial Basis Functions method (Thin Plate Spline model) was also found as the
most unsuitable method for estimation of these elements (MAE=0.1757, MBE=-0.0046 and RMSE=0.261).

Table 2. The best-fitted semivariogram models and their parameters for pH and EC
Soil properties model nugget sill nugget/sill R
2

pH spherical 0.021 0.047 0.562 0.983
EC(Log tranformation) spherical 0.001 2.069 0.999 0.983

In order to understand spatial variation of pH, a map was provided by kriging method for agricultural fields
of Golestan province (Figure 5). Based on this map, the west to east of Golestan province had different
concenrations of pH. In these areas, pH was less than 7.65, and lower than the mean pH in other agricultural lands.
In contrast, the central area was a high pH zone (from 7.65 to 8.25).


Figure 1. Location of the study area

Figure 2. Distribution of sampling points in the province extent

0
106
212
318
0.15 17.37 34.58 51.80
F
r
e
q
u
e
n
c
y
EC
NonTransformed
0
29
57
86
-1.90 -0.44 1.03 2.49 3.95
F
r
e
q
u
e
n
c
y
ln(EC)
Transformed

Figure 3. Histograms of raw and log-transformed values of EC

The semivariogram of the EC data was also spherical with nugget, and sill, nugget/sill equal to 0.001,
2.069, and 0.999, respectively (Table 1). The nugget/sill ratio of EC was 99.9% and belonged to the scope of weak
spatial dependence (Figure 6). Commonly, weak spatial dependence can be recognized to extrinsic factors such as
industrial production, fertilization and other soil management practices; the anthropogenic factors changed their
spatial correlation after a long process of utilization (Cambardella et al., 1994; Shi et al., 2007).
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

163



Figure 4. Empirical and fitted Spherical semivariogram for the pH
Table 3. Results of geostatistical analyses of pH in agricultural lands of Golestan province, north of Iran














In this study, kriging, IDW and RBF were used to estimate electrical conductivity (EC). The summary
statistics for geostatistic method show that kriging (Spherical model) provides much better estimation results for EC
than other methods. Kriging is a widely used method of geostatistical interpolation that assumes that no regional
trend exists in the data. Comparison between the different methods was carried out by MAE, MBE, and RMSE
statistical parameters. In this research, Radial Basis Functions method (Thin Plate Spline model) was found as the
most unsuitable method for estimation of EC (Table 4).


Figure 5. Interpolation map of pH produced by kriging

MAE MBE RMSE Model Method
0.1223 0.0010 0.1745 Gussian
0.1224 0.0012 0.1747 Exponential Kriging
0.1220 0.0006 0.1745 Spherical
0.1420 0.0007 0.1998 Multiquadric
0.1243 0.0064 0.1757 Invers Multiquadric Radial Basis Function(RBF)
0.1757 0.0046 - 0.2601 Thin Plate Spline
0.1256 0.0056 0.1769 Power 1
0.1309 0.0056 0.1845 Power 2 Inverse Distance
Weighted(IDW)
0.1373 0.0060 0.1944 Power 3

Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

164

Spatial patterns of EC estimated by kriging are shown in Figure 7. The results show the high dependence
of this parameter on climate. Electrical conductivity is higher in the northern, semi-arid zone, than in the south and
west. Most agricultural lands of Golestan had EC of saturated soil extract less than 2.75 dS/m, only a small area is
with high EC .

Table 4. Results of geostatistical analyses of EC in agricultural lands of Golestan province, north of Iran



















Figure 6. Empirical and fitted Spherical semivariograms for the log-transformed EC

Figure. 7. Interpolation map of EC produced by kriging
MAE MBE RMSE Model Method
2.0153 0.0052 - 4.6477 Gussian
1.9879 0.0083 - 4.5772 Exponential Kriging
1.9772 0.0066 - 4.5063 Spherical
2.0661 0.0668 4.6041 Multiquadric
1.9085 0.1605 - 4.6430 Invers Multiquadric Radial Basis
Function(RBF)
2.4193 0.1423 - 5.6611 Thin Plate Spline
1.9095 -0.2050 4.5321 Power 1
1.9731 0.1280 - 4.5900 Power 2 Inverse Distance
Weighted(IDW)
1.9322 -0.0865 4.5703 Power 3

Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

163


Figure 8. Map of rain in agricultural fields

The results showed that kriging was the best method and Radial Basis Functions method was also found
as the most unsuitable method to estimate pH. Maroofi et al., (2009) findings are in agreement with our result in
which these researchers reported that Radial Basis Functions were the most unsuitable method for estimating EC
and pH in stream drained water in Hamedan-Bahar plain, west of Iran. Spatial interpolation technique such as
kriging utilized the coregionalization structure of soil properties and provided unbiased estimates and minimum
variance (Ali and Malik, 2010). Based on pH map, the west to east of Golestan province had different
concentrations of pH. The results also can show the high dependence of this parameter on climate such as: rainfall
and agricultural management. Generally, in semi-dried and dried regions, the low rainfall was mainly due to the
high pH. Figure 8 presents Map of rain in agricultural fields.
In Comparison of RBF, IDW and Kriging methods, that kriging (Spherical model) provides much better
estimation results for EC than other methods. Generally, geostatistics is superior to IDW which is similar to the
results of Ahmad (2002), Barca and Passarella (2007), Mashayekhi et al., (2007); Liu et al., (2004), Ali and Malik
(2011) and Taghizadeh Mehrjardi et al., (2008). In an earlier study in Iran, Mashayekhi et al., (2007) stated that
kriging and cokriging were the best methods to estimate EC and SAR in Lenjan-Isfahan plain. Also, in agricultural
lands of Rhodope District, northeastern Greece, Pisinaras et al., (2010) constructed the spatial distribution of soil
EC using ordinary kriging.
Spatial patterns of EC estimated by kriging show the high dependence of this parameter on climate.
Rainfall may be associated with these results (Figure 8), as in the south the water can leach soils and salts are not
presented in the upper horizons in high concentrations. Water movement into soils is the key factor in management
of salt-affected soils. Infiltration rates (IR) and hydraulic conductivity (K) decrease with decreasing soil salinity and
with increasing exchangeable sodium. Infiltration rates are more strongly affected by low salinity and exchangeable
sodium levels than are hydraulic conductivities because of the mechanical impact and stirring action of the applied
water and the freedom for soil particle movement at the soil surface (Oster et al. 1996; Jordan et al. 2004). In
addition, the variability of soil salinity could be high in regions with a great variety of microclimate (Harrach and
Nemeth, 1982). Note that Golestan is a region where the climate changes from the south to the north and thus
rainfall varies from 250 mm to 700 mm per year. It is clear, therefore, that analysis of soil salinity is essential for
understanding the environmental degradation processes in the province. The mitigation and control of soil salinity is
one of the main challenges in the agriculture, particularly where irrigation is used. Meeting this challenge requires
the efficient and accurate quantifying, inventorying, mapping, and monitoring of soil salinity (Pisinaras et al., 2010).

CONCLUSION

This study revealed the spatial variability of available soil EC and pH in agricultural fields of Golestan
province. Results of this research indicated that geostatistics are more suitable methods for estimation of soil
properties than other interpolation methods. The Spherical model is found to be the best model representing the
spatial variability of semivariograms. In this research, the nugget/sill ratio of EC belonged to the scope of weak
spatial dependence. The comparison between the different methods showed that kriging was the best method to
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

166

estimate pH and EC. Based on spatial variability of map, electrical conductivity is higher in the northwest and north
of Golestan (semi arid zone) than in the south and east areas. Almost all the agricultural fields of these areas are
faced with a lot of salinity related problems and the lessening and control of soil salinity is one of the main
challenges in the agriculture of Golestan. It is suggested that in the future studies, other interpolation methods such
as co-kriging and soil properties such as Na and SAR be used in order to prepare precision maps.

ACKNOWLEDGMENTS

We thank the Agricultural and Natural Resources Center of Golestan province and Tarbiat Modares
University (TMU) that supported this research. Also, we sincerely thank Dr A. Fath Nia for his kind helps.

REFERENCES

Ahmed S. 2002. Groundwater monitoring network design: Application of geostatistics with a few case studies from a granitic aquifer in a
semiarid region. In: Sherif MM (ed) Groundwater Hydrology. Balkema, Tokyo, Japan.2: 37-57.
Ali SM, Malik RN. 2011. Spatial distribution of metals in top soils of Islamabad city, Pakistan. Environ Monit Assess .172:116.
Banton O, Seguin MK, Cimon MA.1997. Mapping field-scale physical properties of soil with electrical resistivity. Soil Sci Soc Am J. 61:1010.
Barca E, Passarella G. 2007. Spatial evaluation of the risk of groundwater quality degradation. A comparison between disjunctive kriging and
geostatistical simulation. Environ Monit Assess.137:261-73.
Bishop C. 1995. Neural Networks for Pattern Recognition. Oxford University Press. UK.
Brady NC, Wells RR. 2001.The Nature and Properties of Soil .Upper Saddle River, NJ: Prentice Hall.
Burrough PA, McDonnell RA. 1998. Creating continuous surfaces from point data. In: Burrough PA et al. (eds.) Principles of Geographic
Information Systems. Oxford University Press. UK.
Cambardella CA, Moorman TB, Novak JM, Parkin TB, Turco RF, Konopka AE. 1994. Field-scale variability of soil properties in central lowa
soils. Soil Sci Soc Am J 58:15011511.
Castrignano A, Buttafuoco G, Puddu R. 2008. Multi-scale assessment of the risk of soil salinization in an area of south-eastern Sardinia (Italy).
Precis Agric 9:1731.
Chang YH, Scrimshaw MD, Emmerson RH, Lester JN. 1998. Geostatistical analysis of sampling uncertainty at the Tollesbury managed retreat
site in Blackwater Estuary, Essex, UK: kriging and cokriging approach to minimise sampling density. Sci Total Environ\ 221:4357.
Chien YL, Lee DY, Guo HY, Houng KH. 1997. Geostatistical analysis of soil properties of mid-west Taiwan soils. Soil Sci 162:291297.
Corwin DL, Lesch SM, Shouse PJ, Soppe R, Ayers JE. 2003. Identifying soil properties that influence cotton yield using soil sampling directed
by apparent soil electrical conductivity. Agron J 95:352364.
Corwin DL, Lesch SM. 2003. Application of soil electrical conductivity to precision agriculture: Theory, principles, and guidelines. Agron J
95:455471.
Corwin DL, Lesch SM. 2005. Characterizing soil spatial variability with apparent soil electrical conductivity:I. Survey protocols. Comput Electron
Agr 46.103133
Cressie C. 1990. The origins of kriging. Math Geol 22:239252.
Deutsch CV, Journel AG. 1998. GSLIB: Geostatistical software library and users guide. Oxford University Press, Oxford, UK.
Gawel LJ. 2009. A guide for remediation of salt/hydrocarbon impacted soil. Bismarck, ND: North Dakota Industrial Commission, Department of
Mineral Resources.
Goovaerts P. 1999. Geostatistics in soil science. Geoderma 89:145.
Gundogdu KS Guney I. 2007. Spatial analyses of groundwater levels using universal kriging. J Earth Sys Sci 116 (1): 49-55.
Haykin S. 1994. Neural networks: a comprehensive foundation. Macmilln College Publ. Compny, New York.
Istok JD, Cooper RM. 1998. Geostatistics applied to groundwater pollution. III: global estimates. J Environ Eng 114: 915-928.
Jordan MM, Navarro-Pedreno J, Garca-Sanchez J, Mateu P. 2004. Spatial dynamics of soil salinity under arid and semi-arid conditions:
geological and environmental implications. Environ Geol 45:448456
Juan P, Mateu J, Jordan MM, Mataix J, Melendez-Pastor I, Navaro Pedreno J. 2010. Geostatistical methods to identify and map spatial
variations of soilsalinity. J Geo Exp 46:103-133.
Krige, DG. 1951. A statistical approach to some basic mine valuation problems on the Witwatersrand. J Chemical, Metallurgical and Mining Soc.
South Africa 52:119 139.
Leuangthong O, McLennan JA, Deutsch CV. 2004. Minimum acceptance criteria for geostatistical realizations. Nat Resour Res 13:131141.
Lin GF, Chen LH. 2004. A spatial interpolation method based on radial basis function networks incorporating a semivariogram model. J Hydro
288: 288298.
Liu XM, Xu JM, Zhang MK, Huang JH, Shi JC, Yu X. 2004. Application of geostatistics and GIS technique to characterize spatial variabilities of
bioavailable micronutrients in paddy soils. Environ Geol 46:189194.
Maroofi S, Toranjeyan A, Zare Abyaneh H. 2009. Evaluation of geostatistical methods for estimating electrical conductivity and pH of stream
drained water in Hamedan-Bahar Plain. J Water and Soil Conser 16: 169-187. (In Persian).
Mashayekhi K, Asadi Z, Movahedi Naeini SA, Hajrasuliha S. 2007. Salinity regionalization with geostatistic method in a wet soil in southern
Lenjan-Isfahan (Iran). Indian J Agric Res 41:1 -9.
Mueller TG, Hartsock NJ, Stombaugh TS, Shearer SA, Cornelius PL, Barnhise RI. 2003. Soil electrical conductivity map variability in limestone
soil overlain by loess. Agron J 95:496507.
Oster JD, Shainberg I, Abrol IP. 1996. Reclamation of salt-affected soil. In: Agassi M. (ed.) Soil erosion, conservation and rehabilitation. Marcel
Dekkar, New York.
Pisinaras V, Tsihrintzis VA, Petalas C, Ouzounis K. 2010. Soil salinization in the agricultural lands of Rhodope District, northeastern Greece.
Environ Monit Assess 166:7994.
Pozdnyakova L, Zhang R. 1999. Geostatistical analyses of soil salinity in a large field. Precision Agric 1:153-165.
Intl J Agri Crop Sci. Vol., 4 (4), 157-167, 2012

167

Robinson TP, Metternicht G. 2006. Testing the performance of spatial interpolation techniques for mapping soil properties. Comput Electron Agri
50: 97-108.
Shi J, Wang H, Xu J, Wu J, Liu X, Zhu H, Yu C. 2007. Spatial distribution of heavy metals in soils: a case study of Changxing China. Environ
Geol 52:110.
Sun Y X, Wu CZ, Zhu K, Cui Z, Chen X, Zhang F. 2009. Influence of interpolation method and sampling number on spatial prediction accuracy
of soil Olsen-P. Chinese J Applied Ecol 20:673-678.
Utset A, Ruiz ME, Herrera J, Leon DP .1998. A geostatistical method for soil salinity sample site spacing. Geoderma 86:143151.
Uyan M, Cay T. 2010. Geostatistical methods for mapping grounwater nitrate concentrations. Paper presented at the 3
rd
international conference
on cartography and GIS. Nessebar, Bulgaria. 15-20 June 2010.
Webb A, Shannon S. 1998. Shape-adaptive radial basis functions. IEEE Transactions on Neural Networks 9:1155-1166.
Webster R, Oliver MA. 2001. Geostatistics for environmental scientists. Wiley, New York.
Webster R. 1994. The development of Pedometrics. Geoderma 62:1 15.
Yu N, Zhang Y, Huang Y, Zou H, Ji J, Bai Y, Meng Q. 2008. Micro-scale spatial variability of surface soil pH under different irrigation methods in
greenhouse. Soils 5:5-28.
Zhang C, McGrath D. 2004. Geostatistical and GIS analysis on soil organic carbon concentrations in grassland of southeastern Ireland from two
different periods. Geoderma 119: 261-275.
Zhang CS, Selinus O, Schedin J. 1998. Statistical analyses on heavy metal contents in till and root samples in an area of southeastern Sweden.
Sci Total Environ 212: 217 232.
Zhang CS, Selinus O, Wong P. 2000. Spatial structures of cobalt, lead, and zinc contents in tills in southeastern Sweden. GFF (Transactions of
the Geological Society in Stockholm) 122. 2:213217.
Zhu Q, Lin HS. 2010. Comparing ordinary kriging and regression kriging for soil properties in contrasting landscapes. Pedosphere 20:594606.