Analysis of the Characteristics and Cause Analysis of Soil Salt Space Based on the Basin Scale

Abstract

The analysis of water-soluble salts in the soil is an important basic work for the development and research of the salinity monitoring and salinization control of saline soils, aiming at the complexity of the development of soil salinization in oases in arid and semi-arid areas. Based on the regionalization theory, GPS positioning technology was adopted in this paper to conduct fixed point sampling of the Weigan River Basin oasis from April to May 2020. Soil sampling levels were between 0 m and 0.25 m, between 1 m and 2 m, and between 2 m and 3 m, respectively, and soil physical and chemical properties were analyzed and tested from May to June 2022. The spatial variability of soil salinity at different depths in the Weigan River Basin oasis was quantitatively studied by GIS and geostatistics, and the results show that: (1) the soil salinity of each layer in the Weigan River Basin oasis was generally high, with an average value of 1.27%, 0.87%, and 0.79%, respectively, and a variation coefficient between 1.023 and 1.265, showing strong variability. (2) A reasonable number of soil sampling should be based on the 95% confidence level and 20% relative error. Therefore, the number of soil samples in the corresponding layers should not be less than 89, 70, and 91. (3) Optimal fitting models of soil salinity at the layers between 0 m and 0.25 m and between 1 m and 2 m were both spherical models, while the optimal fitting model of soil salinity at the layer between 2 m and 3 m was an exponential model. Soil salinity at different depths is affected by random factors and structural factors. Soil salinity at the layer between 0 m and 0.25 m showed moderate spatial correlation, while soil salinity at the layers between 1 m and 2 m and between 2 m and 3 m showed strong spatial correlation. (4) Three layers of soil salinity had obvious anisotropy, and the maximum variation direction was northwest–southeast. The nested model can overcome the influence of directionality on the fitting of soil salinity variation function in the layer between 0 m and 0.25 m and improve the spatial interpolation accuracy. (5) In terms of spatial pattern, the area of high soil salinity was located along the Tarim River and at the edge of oasis (i.e., desert–oasis ecotone). On the whole, the salinity shows a gradual increasing trend from the inner part of the oasis (i.e., cultivation area) to the edge of the oasis, and from the northwest to the southeast. (6) The spatial distribution pattern of soil salinity in the study area was mainly affected by the differences in topography, groundwater level, and water quality, while agricultural activities intensified the formation of the distribution pattern, reflecting the complexity of the development of soil physical and chemical properties in arid and semi-arid areas. This study provided a reference for the improvement, management, and rational utilization of saline soil in the Weigan River Basin oasis.

1. Introduction

Soil, as a natural carrier of history, is influenced by climate, topography, irrigation, fertilization, groundwater environment, and other factors, and has complexity and spatio-temporal variability, and its variability with spatial location is known as the spatial variability of soil [1]. This property of soils belongs to the regionalized variables, which differ from purely random variables, but are specific values of random variables at defined locations in the domain and are location-dependent random functions [2]. For this property of soils, geostatistics has long proved to be one of the most appropriate methods to solve the above problem, as it reveals the anomalies and correlations of soil properties on certain scales [3]. To some extent, the spatial variation in soil salinity reflects the degree and state of soil salinization in the soil tillage layer. Understanding its variation is of great significance for guiding people to carry out irrigation and drainage according to the law of variation and the dynamics of soil salinity, formulating measures to prevent and control soil salinization, ensuring soil quality, and improving agricultural output [4,5,6].

In recent years, more and more scholars have studied the spatial variability of soil salinity. Emadi et al. studied the law of distribution of soil pH and salinity in the northern coastal area of Iran by geostatistics [7]; Taghizadeh-mehrjardi et al. studied the spatial variation characteristics of soil salinity in the Ardakan region of central Iran [8]; Douaik et al. analyzed the spatial variation of soil salinity in the field of eastern Hungary [9]; Said Eljebri studied the spatial distribution of properties of soil salinity in the Doukkala irrigated plain of Morocco [10]; Yu studied the effect of rainfall on spatial variability in soil salinity in the coastal of eastern China [11]; and Abderraouf Benslama analyzed the spatial variation in soil salinity of palm groves in the Algerian plain [12]. Most of the above studies focused on coastal and plain areas, and the research objects were mainly field scale, with a relatively small portion of the research on the spatial heterogeneity of soil salinity in arid areas. With the development of oasis-intensive agriculture in the Weigan River Basin, people’s management of farmland has been gradually improved, and irrigation and drainage facilities have been continuously improved, so that salinization in the study area has been mitigated and salt has been gradually shifted from the inner part (i.e., cultivation area) to the edge of the oasis, which has further aggravated the degree of salinization of the lower reaches of the basin and the oasis edge (i.e., desert–oasis ecotone). Its development has seriously threatened human production and life, as well as the ecological security of the basin [13,14]. Therefore, soil salinization control has become important in managing the oasis.

The scientific governance and utilization of soil salinity is based on a correct understanding of the spatial variability of regional soil properties [15,16]. Therefore, in order to avoid further soil degradation, ensure sustainable development of agriculture, and maintain ecological security of the basin, it is necessary to analyze the spatial variation characteristics of soil salinity in the Weigan River Basin oasis. Through field sampling and indoor testing, the spatial variation properties of soil salinity in the layers between 0 m and 0.25 m, between 1 m and 2 m, and between 2 m and 3 m in the Weigan River Basin oasis were quantitatively studied using the methods of traditional statistics, geostatistics, and GIS. The specific objectives include: (1) descriptive analysis of soil salinity. Traditional statistics were used to summarize the overall situation of soil salinity and assess the rationality of the number of samples. (2) Estimation of spatial variability. The spatial heterogeneity of soil salinity at different depths was analyzed using geostatistical theories and methods. (3) Digital mapping of soil salinity. On the basis of spatial variation analysis, the spatial distribution map of soil salinity was drawn based on nested theory and GIS technology. (4) Analysis of the causes of spatial variability. Reasonable control measures were proposed in the discussion of the natural and human factors affecting the spatial variation of soil salinity in this area.

2. Materials and Methods

2.1. Overview of the Study Area

The Weigan River Basin oasis is located in the south of the Queletag Mountain and the middle of the northern foot of the Tarim Basin, a typical and complete pre-mountain alluvial fan plain in southern Xinjiang. It belongs to Aksu region, Xinjiang, China, and encompassing Kuqe, Xinhe, and Shaya Counties within its jurisdiction. Its geographic coordinates: 81°59′37″ E~84°08′29″ E, 40°45′31″ N~41°57′37″ N [13]. The main rivers flowing through the oasis are the Weigan River, the Yindaya River, and the Tarim River, with the Weigan and Yingdaya Rivers being the main sources of the oasis’ irrigation water. The study area has a temperate continental arid climate with severely cold winters and hot summers, large annual and daily temperature differences, and evaporative precipitation ratios of up to 26.4 [17]. The main crops in the oasis are wheat, maize, cotton, walnuts, and dates, and the dominant vegetation in the desert area are halostachys caspica, kalidium foliatum, halocnemum strobilaceum, tamarix, and Populus euphratica.

There are salt spots on the surface of flat areas. Generally, after reclamation, due to unscientific irrigation, the groundwater level rises, and salinization increases. According to soil survey data, the salinized area has reached more than 50% of the existing cultivated land, of which 30% has been severely salinized [18]. Therefore, soil salinization is one of the biggest problems hampering the development of oasis agriculture in this region. The water quality change of oasis water resources in the Weigan River Basin is a process from mild to saline. An extremely arid continental climate, strata and soils saturated with soluble salts, and Meso–Cenozoic salt-bearing areas exposed in the low mountain areas all created favorable conditions for the water mineralization process [17]. The salinization landscape is shown in Figure 1.

2.2. Sample Collection and Test

In this study, the satellite image data of Weigan River Basin oasis in April 2018 is used as the reference map, and the precision of remote sensing image is 30 m. Sampling points are arranged in Weigan River Basin oasis through GIS to cover the whole oasis, and field sampling is conducted from 5 April to 5 May 2020. No large-scale irrigation activities were carried out in the Weigan River Basin oasis during this period, and the evaporation was strong; therefore, it was a typical salinization period in this area and was representative. In the sampling process, the soil texture and vegetation type in the study area are fully considered, and the sample points arranged are appropriately modified according to the actual land use conditions in the study area, so as to achieve uniform sampling and facilitate geostatistical analysis. A total of 92 sampling points is arranged (Figure 2). In order to reduce the soil difference, five sampling points are arranged with the sampling point as the center and 10 m as the radius to form a quincunx distribution. The hand-held undisturbed sampling drill (Model: HW-T30, drill diameter: 38 mm) is used to carry out sampling in the layers between 0 m and 0.25 m, between 1 m and 2 m, and between 2 m and 3 m. The corresponding layers of soil at the five sampling points are mixed and about 0.5 kg soil samples are taken back to the laboratory by quartering method.

2.3. Sample Processing and Analysis

The collected soil samples shall be taken back to the laboratory for natural air drying and grinding, and then passed through a 2 mm sieve for standby. With reference to the agricultural chemical analysis method of soil [19], they shall be filtered and extracted at a water soil ratio of 5:1, and the total salt content in the leaching solution shall be determined. Descriptive statistical analysis of test samples was completed by SPSS; Use GS+ to calculate the semi variance function and select the theoretical model; Spatial interpolation and anisotropic nesting are completed by geostatistical analyst module in ArcGIS; Use Origin and CorelDraw to draw corresponding maps.

2.4. Research Methods

(1)

Semi variogram

As a basic function to study the spatial variation of soil attributes, some important parameters such as range (a), nugget(C₀), sill(C₀ + C), and fractal dimension can reflect the variation degree and autocorrelation range of soil characteristics on a certain scale [5]. The calculation formula is shown in Equation (1).

$$\gamma (h)=\frac{1}{2N(h)}{\displaystyle \sum }_{i=1}^{N(h)}{[Z(x_i)-Z(x_i+h)]}^2$$

(1)

where $\gamma \left(h\right)$ is the experimental variogram; N(h) is the number of sample pairs; h is the vector between the two sample points; Z(x_i) and Z (x_i + h) are the observed values of Z(x_i) at spatial positions x_i and x_i + h, respectively.

(2)

Estimation of reasonable sampling number

For the estimation of the number of samples, the commonly used formula is the sampling number calculation formula constructed by Cochran for regional pure random sampling [20].

$$n={\left(t\times Std\right)}^2/d^2$$

(2)

where n is the reasonable number of samples, t is the deviation corresponding to the significance level of ∝ in the distribution table, Std is the sample standard deviation, and d is the average value of the sample × relative error (%).

(3)

Anisotropic nesting

For most geometric anisotropy, the directional variation diagram approximates an ellipse [21]. At this time, let a₁ be the major axis of the ellipse, which coincides with the horizontal direction, a₂ be the minor axis, which coincides with the vertical direction (Figure 3), and k = a₁/a₂, then the linear transformation matrix is: A = $\left[\begin{array}{cc}1& 0\\ 0& k\end{array}\right]$, the transformed $h^{\prime }$ is shown by Equation (3).

$$h^{\prime }=\left[\begin{array}{c}{h}_w^{\text{′}}\\ h_v^{\text{′}}\end{array}\right]=\left[\begin{array}{cc}1& 0\\ 0& k\end{array}\right]=\left[\begin{array}{c}{h}_w\\ h_v\end{array}\right]=\left[\begin{array}{c}{h}_w\\ k{h}_v\end{array}\right]$$

(3)

At this time, the pattern of the shape of the ellipse becomes a circle with the major axis as the radius (Figure 3), and the anisotropy becomes isotropy. When a spherical model is used to fit the vertical variation, it can be written using Equation (3).

$$\gamma \left(h^{\prime }\right)=C\left(\frac{3h^{\prime }}{2a_1}-\frac{{h^{\prime }}^3}{2a_1^3}\right)=C[\frac{3kh}{2a_1}-\frac{{\left(kh\right)}^3}{2a_1^3}]$$

(4)

The variation in the horizontal direction is γ(h), and the final anisotropic nesting result is expressed in Equation (4).

$${\gamma }^{*}\left(h\right)=\gamma \left(h^{\prime }\right)+\gamma \left(h\right)$$

(5)

In practical research, there is no specific variation direction (Figure 4), which is generally studied by first rotating the coordinate axes by an angle φ so that they are parallel to the major axis of the ellipse, and then converted into Isotropy through a linear transformation in Equation (3).

(4)

Model inspection

I-value test method and cross-validation method are used to test the optimality of the theoretical semi-variogram model, which is the calculation method as follows [22,23]:

$$I=\overline{\widehat{[Z}\left(x_i\right)-Z\left(x_i\right){]}^2}\times \left[P\times \left|1-\frac{1}{\overline{{[\frac{\widehat{Z}\left(x_i\right)-Z\left(x_i\right)}{\widehat{\mathsf{\sigma }}\left(x_i\right)}]}^2}}\right|+\left(1-P\right)\right]$$

(6)

$$P=\{\begin{array}{c}0.1, 0\le \overline{\widehat{[Z}\left(x_i\right)-Z\left(x_i\right){]}^2 }\le 100 \\ 0.2, 100\le \overline{\widehat{[Z}\left(x_i\right)-Z\left(x_i\right){]}^2 } \end{array}$$

(7)

where $\widehat{\sigma }\left(x_i\right)$ is the standard error of Kriging prediction, $\widehat{Z}\left(x_i\right)$ is the predicted value of the ith sample at position x, $Z\left(x_i\right)$ is the measured value of the ith sample at position x, and P is an empirical parameter. The smaller the value of I, the better the representativeness of the theoretical model to the spatial variation structure of the soil.

3. Results and Analysis

3.1. Basic Statistical Analysis of Soil Salt Content

In order to understand the overall variation of soil salinity in the study area and observe whether the distribution law of the measured sample data meets the normal distribution, descriptive statistical analysis should be conducted on the data of the sampling points in the study area before geostatistical analysis. The statistical results are shown in

Table 1. Descriptive statistical analysis of soil salinity.

Depth/m	n	Min	Max	Mean	SD	Skewness	Kurtosis	CV	Converted p-Value
		(%)	(%)	(%)
0–0.25 m	92	0.04	10.35	1.27	1.46	2.72	7.74	1.150	0.2
1–2 m	92	0.01	5.64	0.87	0.89	2.56	7.23	1.023	0.2
2–3 m	92	0.06	5.11	0.79	0.92	2.73	6.46	1.265	0.052

Note: n is the number of samples; SD is the standard deviation; CV is the coefficient of variation.

.

It can be seen from Table 1 that the average soil salinity of layers between 0 m and 0.25 m, between 1 m and 2 m, and between 2 m and 3 m was 1.27%, 0.87%, and 0.79%, respectively. According to the classification standard of saline soil in Xinjiang [24], the layer between 0 m and 0.25 m was severely salinized, while the layers between 1 m and 2 m and between 2 m and 3 m were moderately salinized. The coefficient of variation of soil salinity in the layers between 0 m and 0.25 m, between 1 m and 2 m, and between 2 m and 3 m was 1.59, 1.07, and 1.34, respectively, showing a trend of first decreasing and then increasing. According to the classification of the coefficient of variation [3], the coefficient of variation of the three layers maintained strong variability. At the same time, the difference between the maximum value and the minimum value of soil salinity of the three layers was large, indicating that the horizontal distribution of soil salinity was uneven. The reasons are as follows: (1) the evaporative precipitation ratios is large in the arid area, and the groundwater rises to the surface in the form of capillary water under the action of water potential gradient, and under the strong evaporation, the soil salt “surface accumulation” phenomenon is caused. (2) Due to the poor anti-seepage effect of the canal system, 59% of the water is lost in the process of water conveyance, which raises the groundwater level and promotes the formation of salinization. (3) The imbalance of diversion and drainage, or “only irrigation without drainage” and “upper drainage and lower irrigation” accelerate soil salinization. Through a K–S test, it is found that p values are all 0, which does not conform to normal distribution. Therefore, it is necessary to take the logarithm of the measured sample data to eliminate the scaling effect. The frequency distribution after conversion is shown in Figure 5, and its significance is greater than 0.05, which conforms to normal distribution.

3.2. Accuracy Analysis of Sampling Number

In order to assess the expected value of each parameter with a limited number of observations and to ensure sufficient confidence, a reasonable number of samples must be determined [25]. The spatial distribution of soil properties has its inevitability and chance [26], and this experiment focuses on the chance study of the spatial distribution of soil properties. According to Cochran’s formula, the number of samples needed for the whole watershed at a certain confidence level and relative error can be found, and the calculation results are shown in

Table 2. Statistics of sampling number of soil salt content.

Depth/m	95% Confidence Level			95% Confidence Level
	5%	10%	20%	5%	10%	20%
0–0.25	1422	355	89	866	217	54
1–2	1126	281	70	686	171	43
2–3	1459	365	91	889	222	56

.

It can be seen from Table 2 that with the increase in soil depth, soil variability decreases first and then increases, and the number of samples required to reach the same confidence level and error also decreases first and then increases. With the reduction in confidence level and relative error requirements, the number of samples required for each layer of soil is also greatly reduced. Cochran’s research indicates that the relative error of reasonable sampling should be determined according to the actual level and critical level of soil salinity. In this study, average soil salinity of layers between 0 m and 0.25 m is 1.27%, when the corresponding errors are 5%, 10%, and 20%, the fluctuation range of soil salinity is 1.21–1.33%, 1.14–1.40%, and 1.02–1.53%. Average soil salinity of layers between 1 m and 2 m is 0.87%, and the fluctuation range of corresponding error is 0.83–0.91%, 0.78–0.96%, and 0.69–1.04%. Average soil salinity of layers between 2 m and 3 m is 0.79%, and the fluctuation range of corresponding relative error is 0.75–0.83%, 0.71–0.87%, and 0.63–0.95%. According to the salinization classification standard, the fluctuation range of soil salinity in the layers between 0 m and 0.25 m under the three error levels is within the range of severely salinized, and the fluctuation range of soil salinity in the layers between 1 m and 2 m, and between 2 m and 3 m under the three error levels is basically within the range of moderately salinized. Therefore, a reasonable choice of sampling number in terms of economy and accuracy as well as sampling efficiency was made, and the reasonable sampling numbers of soil salinity in the layers between 0 m and 0.25 m, between 1 m and 2 m, and between 2 m and 3 m were 89, 70, and 91, respectively, and the actual sampling number met the requirement of reasonable sampling number under this condition and was suitable for variability analysis.

3.3. Spatial Variation Characteristics of Soil Salinity

3.3.1. Isotropic Analysis

The traditional statistical analysis of soil salinity can only summarize the whole picture of soil salinity, but it is difficult to express the randomness and structure of soil salinity. Therefore, it is necessary to analyze and discuss its spatial variation structure using geostatistics theory and methods [4]. The selection of the optimal semi variance function model generally follows the principle that the larger the determination coefficient (R²), the smaller the residual error (RSS) [16]. See

Table 3. Variogram model and parameters.

Depth/m	Model	Nugget C0	Sill C0 + C	Range a/km	Nugget Coefficients C0/(C0 + C)	R2	RSS
0–0.25	Spherical	1.459	2.086	56.0	69.5	0.756	0.108
1–2	Spherical	0.010	1.216	18.2	0.8	0.830	0.042
2–3	Exponential	0.139	1.309	24.9	10.2	0.761	0.129

and Figure 6 for the results.

It can be seen from Table 3 that the fitting models of the salinity variation function of the three layers of soil were the spherical model, spherical model, and exponential model, respectively, and the determination coefficients (R²) of the theoretical model to the variogram were 0.756, 0.830, and 0.761, respectively, indicating that the theoretical model had a high degree of fitting with the actual variation function. In the semi-variance function model, the nugget (C₀) represents the variation caused by random factors, including human factors, such as irrigation, cultivation, and management. Partial Sill(C) represents the variation caused by structural factors, including natural factors such as soil parent material, topography, and groundwater level during soil formation [27,28]. Nugget coefficient [C₀/ (C₀ + C)] can reflect the degree of correlation of variables. Wherein, when C₀/(C₀ + C) < 25%, soil salinity has a strong spatial correlation; when 25% < C₀/(C₀ + C) < 75%, soil salinity has a moderate spatial correlation; and when C₀/(C₀ + C) > 75%, soil salinity has a week spatial correlation [29,30]. The nugget coefficient of soil salinity in the layer between 0 m and 0.25 m was 69.5%, showing moderate spatial correlation, which indicates that the spatial heterogeneity of surface soil salinity is affected by both random factors and structural factors. However, on the scale of the basin as a whole, the surface soil is affected by random factors more than structural factors. The nugget coefficients of soil salinity in the layers between 1 m and 2 m and between 2 m and 3 m were 0.8% and 10.2%, respectively, which is far less than 25%, showing a strong spatial correlation. This ratio indicates that the salinity of deep soils is almost completely spatially autocorrelated under the action of structural factors.

Range(a) reflects the autocorrelation range of the variable [16], and range(a) of the soil salinity in the three layers was 56 km, 18.2 km, and 24.9 km, respectively. The layer between 0 m and 0.25 m had the largest range(a) of soil salinity (56 km), indicating that the soil salt content is consistent in a large range. This is because the study area mainly implements a highly intensive management mode of agricultural technology, so the agricultural management measures in different areas are similar. Long-term continuous farming and management have weakened the influence of structural factors such as topography and groundwater level, resulting in an increase in salinity range(a) in the surface soil [31,32,33]. Meanwhile, salinity in soils of layers between 1 m and 2 m and between 2 m and 3 m are mainly influenced by structural factors such as topography and groundwater level, and the degree of this influence increases with depth, specifically, the range values (a) of soil salinity increased significantly from that of the soil layer between 1 m and 2 m [34].

3.3.2. Anisotropy Analysis

Soil, as a carrier of history, is influenced by natural and anthropogenic factors during its formation and evolution, which leads to a certain directionality in the spatial structure of soil properties [35]. In the analysis of the structure of the variation function, it is most important to analyze the directionality of the variation and find out the maximum direction of variation, which, in turn, provides the necessary parameters for the construction of the nested model [22]. The results are shown in

Table 4. Anisotropic parameters of soil salinity at different depths.

Depth/m	Major Axis Direction Angle/°	Major Axis Range a/km	Minor Axis Range a/km	Anisotropy Ratio/k	Nugget C0	Sill C0 + C
0–0.25 m	133	75.39	44.66	1.69	1.416	2.002
	43	44.66	75.39	0.59	1.416	2.002
1–2 m	120	44.36	15.58	2.84	0.02	1.227
	30	15.58	44.36	0.35	0.02	1.227
2–3 m	124	76.56	25.08	3.05	0.127	1.354
	34	25.08	76.56	0.32	0.127	1.354

.

It can be seen from Table 4 that the soil salinity of layers between 0 m and 0.25 m belongs to geometric anisotropy at 133° and 43° and the maximum anisotropy ratio at 133° is 1.69, indicating that the spatial variability of soil salinity in the layers between 0 m and 0.25 m mainly occurs in the northwest southeast direction. The anisotropy ratio of soil salinity in the layers between 1 m and 2 m, and between 2 m and 3 m is the largest at 120° and 124°, respectively, and the maximum variation direction is northwest–southeast, which is basically consistent with the basin topography and flow field direction, and the maximum anisotropy ratio is greater than the surface layer. The above shows that the deep soil changes strongly in the northwest southeast direction under the influence of structural factors, such as topography and groundwater level, while the soil the layers between 0 and 0.25 m weakened the degree of variability in this direction under the effect of random factors such as tillage and irrigation, and thus there was an isotropic trend [33].

3.3.3. Nested Structure Analysis and Spatial Distribution Characteristics

According to the previous analysis, the spatial variability of soil salt content under anisotropic conditions has strong directionality, and the influence of directionality on the fitting of variation function needs to be considered [21]. Since agricultural production activities mostly act on shallow soil, in this study, a single spherical model is used to perform anisotropic nesting on the spatial structure of surface soil salt content. The collated nested model is shown below:

$$\gamma (h)=\{\begin{array}{l}{C}_{0}h=0\\ C_0+C\left(\frac{3h}{2a}-\frac{h^3}{2a^3}\right)a>h>0\\ C_0+Ch>a\end{array}$$

(8)

$${\gamma }^{*}(h)=\{\begin{array}{l}1.416h=0\\ 1.416+0.586\left[\frac{8.07}{2}\left(\frac{h}{a}\right)-\frac{5.826}{2}{\left(\frac{h}{a}\right)}^3\right]75.39>h>0\\ 2.002h>75.39\end{array}$$

(9)

After fitting the variogram with the nested model, the ordinary Kriging interpolation method is used for spatial estimation of the surface sample data. The interpolation results are shown in Figure 7a It can be seen from

Table 5. Comparison of interpolation accuracy in different spaces.

Interpolation Method	I Value	$\overline{[\widehat{\mathit{Z}}({\mathit{x}}_{\mathit{i}})-\mathit{Z}({\mathit{x}}_{\mathit{i}}){]}^2}$
Ordinary Kriging	2.625	2.725
Nested model method	1.588	1.636

that the I and $\overline{[\widehat{Z}(x_i)-Z(x_i){]}^2}$ values of the nested model method are smaller than those of the ordinary Kriging interpolation method, which indicates that the spatial variation structure of soil salinity can be better expressed based on the nested model under anisotropic conditions and the interpolation accuracy can be improved. The soil salinity of layers between 1 m and 2 m and between 2 m and 3 m is plotted according to the ordinary Kriging interpolation method, and the results are shown in Figure 7b,c.

According to the previous analysis, the surface soil salinity shows strong variability and directionality, indicating that its spatial distribution shape is irregular, as shown in Figure 7. The variation range of soil salinity in the layers between 0 m and 0.25 m is 0.213–1.775%, of which mild and partial moderately salinized are distributed in the north of the study area, i.e., inside the oasis (cultivation area), and extend to the southeast in an irregular ellipse. The oasis edge (i.e., oasis desert ecotone) is dominated by moderately salinized, which is distributed circularly. Part of the heavy salinization is distributed in the northeast of the study area, which is a low-lying basin with sparse vegetation and easy to store salt, the rest is distributed in belt along the Tarim River and at the edge of the oasis. On the whole, salinization presents a trend of gradual aggravation from the center to the edge of the oasis. When considering the continuity of spatial variation, the variation increase in salt content in the northwest southeast is relatively stable and the continuity is the best. The distribution pattern of soil salinity in the layers between 1 m and 2 m, and between 2 m and 3 m is similar, and the center of high and low value is Mosaic distribution, the spatial continuity is not as good as the surface, and the degree of fragmentation is higher. The high value centers of soil salinity in the layers between 1 m and 2 m, and between 2 m and 3 m are mainly distributed in the northeast and southeast of the study area, and the low value centers are mainly distributed in the southwest and northwest of the study area.

4. Discussion

The spatial distribution pattern of soil salinity in the study area is affected by many factors. The differences in topography, groundwater level, and groundwater properties may cause this spatial distribution pattern, and agricultural activities will accelerate the formation of this distribution pattern. The economy in the Weigan River Basin oasis is dominated by agriculture, and the traditional flood irrigation is the main irrigation method. Therefore, irrigation and drainage have an important impact on the soil salt content of the oasis. In the Weigan River Basin oasis, the Weigan River and the Yingdaya river flow through a wide area, and there are large and small irrigation channels and alkali drainage channels distributed around these two rivers (as shown in Figure 8). Therefore, taking the two rivers as the research object, 40 points are selected from the upstream to the downstream of the two rivers, and the sampling points are evenly distributed on both banks of the river. According to the sampling points and Kriging interpolation results, the distribution profile of soil salinity in the layers between 0 m and 0.2 m is drawn, and the soil salinization status in this area is analyzed.

Figure 9 is the profile of total salt content of two rivers. It can be seen from the figure that the soil salt content increases with the trend of the river, showing a distribution of low in the upstream and high in the downstream. Topography is an important factor that affects the soil properties and determines its changes [36,37]; these high salinity areas are located in the lower southeast of the study area and the edge of the oasis, as shown in Figure 10.

Analysis of the reasons the delta oasis in this basin undergoes less precipitation, evaporation, and soil salt accumulation shows that the salinization situation is more serious. During irrigation in the upstream region, due to the higher terrain, the salt accumulates with the water to the middle and lower reaches of the lower terrain, and the salt accumulates the most in the lower reaches of the basin and the edge of the oasis, so the soil salt content is the highest. For the Weigan River Basin oasis, soil salinization is an important reason restricting the development of local agriculture [38]. Due to the serious salinization, a large amount of water is used for salt washing in this area every year. The salt enters the alkali drainage channel and is carried into the downstream area by the alkali drainage channel, which also intensifies the formation of the salt distribution pattern in this area. At the same time, this irrigation method also leads to the rise of the water table, further aggravating the degree of salinization in the area. In a basin, groundwater and surface water are mutually replenished. Soil salinity and groundwater mineralization are related to one another; the higher the salt content, the higher the groundwater salinity [17,35]. Therefore, in groundwater and surface water supplies, each of these factors will also affect the salinity of surface water, and an increased in the main surface water salinity of irrigation water will further aggravate the degree of salinization in this area. In the Weigan River Basin oasis, due to the local long-term irrigation mode and unreasonable methods to suppress soil secondary salinization, the oasis in this area has a tendency to expand outward, but at the same time, the degree of salinization is further deepened [13]. As can be seen from Figure 7, the lower reaches of Weigan River are connected with the middle reaches of Tarim River, and some alkali drainage channels in Weigan River Basin oasis are connected with Tarim River, which inevitably drains soil salt in Weigan River Basin oasis directly into Tarim River through irrigation channels, thus aggravating the degree of soil salinization along the Tarim River basin to a certain extent.

As soil salinization has caused serious impacts on both agricultural production and oasis ecological environment, the management and improvement of salinized arable land and prevention of secondary soil salinization to improve land use efficiency, enhance comprehensive agricultural productivity and improve agricultural production conditions have become important tasks for the current in the Weigan River Basin oasis. The above analysis shows that the salinization of the Weigan River Basin oasis is caused by various factors, so the influence of various factors should be considered comprehensively in the management. Firstly, it is necessary to combine local topography, geology, hydrology, meteorology, soil structure, and other natural factors, and also to develop operable and efficient engineering, agro-technical and biological measures in accordance with the local salinization status. Secondly, we should build an efficient irrigation system and a reliable drainage system to effectively control soil salinity and achieve gradual desalination. Irrigation management should be strengthened and irrigation techniques should be improved according to the principle of soil-water balance. Various management mechanisms should also be established when implementing these measures to ensure the implementation of these measures in order to achieve the effect of comprehensive management.

5. Conclusions

(1)

The average values of soil salinity in each layer of soil in the Weigan River Basin oasis were generally high. The soil of the layer between 0 m and 0.25 m was severely salinized, while the soil of the layers between 1 m and 2 m and between 2 m and 3 m were moderately salinized. The three layers of soil salinity did not conform to a normal distribution, and showed strong variability, with the coefficients of variation ranging from 1.15 to 1.26;

(2)

According to Cochran’s formula for calculating the reasonable random sampling size, at the 95% confidence level and 20% relative error, the sampling numbers of the three layers of soil were 89, 73, and 61, respectively. When calculating the reasonable number of samples, not only the accuracy of the sampling should be taken into account, but also the economy and efficiency of sampling should be considered under the premise of ensuring accuracy;

(3)

The spatial variation direction of soil salinity in each layer was evident, and the long-axis range corresponding to the maximum anisotropy ratio first decreased and then increased as the depth decreased. The spatial variation structure of soil salinity in the layer between 0 m and 0.25 m was fitted with a nested model, and then interpolated with Kriging, and it was found that soil salinity showed a gradually increasing trend from the center to the edge of the oasis;

(4)

The analysis of variation function shows that soil salinity in the layer between 0 m and 0.25 m shows a moderate spatial correlation, while soil salinity in the layers between 1 m and 2 m and between 2 m and 3 m show strong spatial correlations, in which topography and groundwater level were the main structural factors affecting soil spatial variability, and farming and irrigation were the main random factors. Therefore, it is proposed that the comprehensive effect of various factors should be considered in the treatment of salinization, especially natural factors such as the local terrain, soil structure, and groundwater level, to build an efficient irrigation system and a reliable drainage system, and to effectively control soil salinization and realize gradual desalination, thereby achieving the effect of comprehensive treatment of salinized soil and improving land use efficiency.