The Characteristics and Driving Factors of Soil Salinisation in the Irrigated Area on the Southern Bank of the Yellow River in Inner Mongolia: A Assessment of the Donghaixin Irrigation District

Abstract

Soil salinisation is a critical problem in northern China’s arid and semi-arid irrigated regions, posing a substantial impediment to the sustainable advancement of agriculture in these areas. This research utilises the Donghaixin Irrigation District, located on the southern bank of the Yellow River in Inner Mongolia, as a case study. This study examines the spatial distribution and determinants of soil salinisation through macro-environmental variables and micro-ion composition, integrating regression models and groundwater ion characteristics to elucidate the patterns and causes of soil salinisation systematically. The findings demonstrate that soil salinisation in the study region displays notable spatial clustering, with surface water-irrigated regions exhibiting greater salinisation levels than groundwater-irrigated areas. More than 80% of the land exhibits moderate salinity, predominantly characterised by the ions Cl⁻, HCO₃⁻, and SO₄²⁻. The hierarchy of ion concentration variation with escalating soil salinity is as follows: Na⁺ > K⁺ > SO₄²⁻ > Cl⁻ > Mg²⁺ > HCO₃⁻ + CO₃²⁻ > Ca²⁺. The susceptibility of ions to soil salinisation is ordered as follows: Ca²⁺ > Na⁺ > HCO₃⁻ + CO₃²⁻ > Mg²⁺ > K⁺ > Cl⁻ > SO₄²⁻. In contrast to the ordinary least squares (OLS) model, the geographic weighted regression (GWR) model more effectively elucidates the geographical variability of salinity, evidenced by an adjusted R² of 0.68, particularly in high-salinity regions, where it more precisely captures the trend of observed values. Ecological driving elements such as organic matter (OM), pH, groundwater depth (GD), total dissolved solids (TDS), digital elevation model (DEM), normalised difference vegetation index (NDVI), soil moisture (SM), and potential evapotranspiration (PET) govern the distribution of salinisation. In contrast, anthropogenic activities affect the extent of salinisation variation. Piper’s trilinear diagram demonstrates that Na cations mainly characterise groundwater and soil water chemistry. In areas irrigated by surface water, the concentration of SO₄²⁻ is substantially elevated and significantly affected by agricultural practises; conversely, in groundwater-irrigated regions, Cl⁻ and HCO₃⁻ are more concentrated, primarily driven by evaporation and ion exchange mechanisms.

1. Introduction

Soil is a fundamental natural resource vital for human existence. Its significance is evident in supplying water, nutrients, and structural support to plants and its profound role in sustaining ecosystem processes and regulating global environmental conditions [1,2,3]. Soil is essential to ecosystems, playing an irreplaceable role in critical natural processes, including the water and carbon cycles, and serving as the primary safeguard for global food security and ecological equilibrium [4,5]. A robust soil structure may efficiently sustain biodiversity and deliver several ecosystem services, including water purification and climate regulation [6,7]. Nonetheless, the escalation of human activities, particularly the imprudent use of agricultural water and soil resources, has rendered soil salinisation a significant danger to the world’s ecological environment and agricultural sustainability [8,9]. Salinisation, a kind of soil deterioration, considerably diminishes the soil’s production capabilities, resulting in limited crop growth, decreased agricultural product quality, and reduced fertiliser usage efficiency [10]. In China, saline soils are predominantly found in the northern and coastal regions, with little rainfall and high evaporation rates. These regions comprise 4.88% of the nation’s arable land, with a varied and extensive distribution [11]. The prevalence and intricacy of salinisation render it a significant impediment to regional agricultural advancement and a concealed danger to national food security [12,13]. Consequently, thoroughly examining and efficiently addressing saline soils is essential for improving agricultural output, advancing green agriculture, and fostering the harmonious development of ecological settings.

Rengasamy [14] characterises soil salinisation as the accumulation of salts or excessive evaporation that inhibits plant development. Salt accumulation transpires not just in the top layer of soil but may also infiltrate deeper strata, resulting in the formation of deep saline soils. Soil salinisation is a multifaceted natural process influenced by several variables [15], including climate, geology, hydrology, and anthropogenic activities, with inadequate irrigation planning being the primary human-induced component [16]. Soil salinisation is especially prevalent in arid and semi-arid areas [17]. It often manifests as the deposition of salts in the surface or root zone, significantly impairing the soil’s hydrolytic, ion-exchange, and nutrient release capacities, consequently impacting crop development [18,19]. The correlation between soil salinisation and groundwater depth (GD), as well as ion chemical composition, is significant. By integrating total dissolved solids (TDS) and stable isotope data for hydrochemical research, the influence of groundwater features on soil salt buildup across various locations may be elucidated [20,21]. Research by Wang et al. [22] indicates that in the Hetao Plain, soil salinisation mainly results from the interplay of elevated GD, evaporation, and irrigation water sources. Xie et al. [23] investigated the influence of cultivation duration on soil salinity and nutrient concentrations, identifying soil particle distribution and nutrient content as critical determinants of salinisation variability. They recommended enhancement strategies, including straw coverage, straw retention, and the application of organic fertilisers. These enhancement strategies help restore the ecological service functions of the soil, such as improving soil moisture (SM) retention, enhancing soil microbial activity, and so on [24]. Furthermore, the research of Guo et al. [25] in the Yellow River Delta indicated that the primary elements affecting soil salinisation fluctuate considerably over time, with temperature and slope exerting varying influences at distinct periods. The mechanism of soil salinisation is intricate, with primary driving elements comprising climatic conditions, topographical characteristics, groundwater quality, and anthropogenic activity [26]. In arid and semi-arid locations, elevated evapotranspiration is regarded as the primary catalyst for salinisation, while increasing GD, low-lying topography, and inferior irrigation water quality further intensify this phenomenon [27,28]. In addition, irrigated agriculture has expanded by 174% globally since 1950 [29], excessive pesticide use and imprudent land use practises significantly affect soil structure and salt movement [30], and there is a long time lag between land use changes and system responses (such as recharge, streamflow, and water quality). Nonetheless, prior research has predominantly concentrated on individual factors influencing soil salinity, with insufficient systematic investigation into the impact of multiple environmental factors on soil salinisation. Additionally, the efficacy of the geographic weighted regression (GWR) model in forecasting regional soil salinity necessitates further validation.

The study area is in the Donghaixin Irrigation District along the southern bank of the Yellow River in Inner Mongolia. Soil salinisation in this region has intensified in recent years [31]. Salinisation has markedly diminished soil fertility and inflicted enduring adverse effects on local agricultural output and the natural environment. The region has significantly higher evaporation than precipitation [32], resulting in conducive circumstances for salt buildup, while anthropogenic causes such as intensive management and over-irrigation have exacerbated the salinisation issue [33]. Consequently, it is essential to undertake a comprehensive investigation of the attributes and origins of soil salinisation in this area. This study examines natural and anthropogenic factors contributing to salinisation at both micro and macro levels, elucidates spatial distribution patterns, and investigates the impact of soil and groundwater ions on soil salinity under different levels of salinisation. Regression analysis is employed to develop a spatial prediction model for soil salinisation, offering a scientific basis for precise prediction and mitigation of soil salinisation and providing theoretical support for the sustainable agricultural development along the southern bank of the Yellow River in Inner Mongolia.

2. Materials and Methods

2.1. Study Area

The study area is situated in the Donghaixin Irrigation District inside Dalad Banner, Ordos City, Inner Mongolia (40.44° N–40.51° N, 109.86° E–110.01° E); see Figure 1. Its temperate continental climate is characterised by cold, arid winters and hot, humid summers. The mean annual temperature is 6.2 °C, and the mean annual precipitation is 348.3 mm, predominantly falling as showers or heavy rain between June and August. The mean annual evaporation is 2506.3 mm. The average soil pH of the study area is 8.47. The cumulative area of salinised soils in Dalad Banner is 718.34 hm² (square hectometre), constituting 9.23% of the total land surface, with 686.38 hm² of salinised land located in the river plain, or 37.16% of the plain area. The Donghaixin Irrigation District employs both surface water and groundwater irrigation techniques. Surface water is conveyed into the irrigation canals from the Yellow River via pump stations and allocated to the fields, while groundwater is extracted and transferred by pipes to the fields. The irrigation technique in the surface water-irrigated area is flooding irrigation, while the irrigation techniques in the groundwater-irrigated area is drip irrigation. The primary crops cultivated in the region are maize and sunflower. The irrigation quotas for maize and sunflower under flooding irrigation are 300 mm and 180 mm, respectively, while the irrigation quotas under drip irrigation are 220 mm and 140 mm, respectively. In the surface water irrigated area, autumn irrigation for maize is carried out from October to November in the previous year, with an irrigation quota of 150 mm, and spring irrigation for sunflower is carried out in late May, with a quota of 135 mm. There is a significant difference in GD in the study area. The GD in the surface water irrigated area is relatively shallow, ranging from 1 to 3 m, while the GD in the groundwater irrigated area is relatively deep, ranging from 8 to 14 m. The physical and chemical properties of the soil and groundwater in the study area are shown in

Table 1. Soil physical and chemical properties of the study area.

Soil Layer Depth (cm)	Soil Particle Size Distribution (%)			Bulk Density (g/cm3)	Texture	Organic Matter (g/kg)	Salinity (g/kg)
	<0.002 mm	0.002–0.05 mm	>0.05 mm
0–20	38.3	38.4	23.3	1.441	Clay loam	14.8	1.82
20–40	37.2	38.9	23.9	1.446	Clay loam	15.9	2.14
40–60	23.7	41.1	35.2	1.428	Loam	15.2	2.49
60–80	19.6	44.3	36.1	1.42	Loam	14.1	2.50
80–100	14.6	47.5	37.9	1.396	Loam	13.6	2.81

and

Table 2. Chemical characteristics of Yellow River water and groundwater in the study area.

Water Source	TDS (g/L)	pH	Ion Content (mg/L)
			K+	Na+	Ca2+	Mg2+	Cl−	SO42−	HCO3− + CO32−
Yellow River Water	0.37	7.51	0.89	97.11	13.32	6.88	38.34	140.91	71.57
Groundwater (Surface Water Irrigated Area)	1.86	7.88	5.07	388.59	105.04	37.55	223.65	550.59	550.82
Groundwater (Groundwater Irrigated Area)	1.68	7.91	4.16	396.54	77.73	30.28	238.46	328.92	601.00

.

2.2. Sample Collection and Measurement

Soil and groundwater samples were obtained in April 2024. The sampling locations were uniformly allocated across the study area, comprising 80 soil and 30 groundwater sampling points and GD assessments. Surface soil samples (0–20 cm) were obtained with a spade. The soil samples were meticulously combined, and residues, including stubble and gravel, were eliminated before placing the samples into self-sealing bags for examination. Groundwater samples were collected in narrow-mouth polyethene vials, cleaned three times with deionised water and let to drain before use. Approximately 5 litres of water were collected from each sampling location, and the samples were preserved at 4 °C for analysis [34]. GD were assessed via a portable GD meter. Soil samples were examined for eleven indicators [35,36]. Soil salinity was quantified through the water-soluble salt extraction and evaporation technique; pH was assessed by creating a soil suspension with a 1:5 soil-to-water ratio, subsequently measured with a pH meter (Testo 206, Kirchzarten, Germany); organic matter (OM) content was evaluated via the potassium dichromate titration method; chloride ions (Cl⁻) concentration was determined using the argentometric technique; carbonate ions (CO₃²⁻) and bicarbonate ions (HCO₃⁻) levels were ascertained through acid-base titration; sulphate ions (SO₄²⁻) was quantified using the gravimetric method, where the SO₄²⁻ was precipitated as barium sulphate (BaSO₄) and weighed after filtration and drying; potassium ions (K⁺) and sodium ions (Na⁺) ions were determined using a flame photometer (INESA 6400A, Shanghai, China); magnesium ions (Mg²⁺) and calcium ions (Ca²⁺) ions were determined using an atomic absorption spectrophotometer (INESA AA7020, Shanghai, China). Groundwater samples were examined for 10 indicators [37,38]. TDS were assessed through weighing; pH was recorded using a pH meter; K⁺, Ca²⁺, Na⁺, and Mg²⁺ concentrations were determined with an ion chromatograph (Dionex ICS-5000, San Jose, CA, USA); The determination methods for Cl⁻, CO₃²⁻, HCO₃⁻, and SO₄²⁻ in groundwater were the same as those in soil. The categorisation of soil salinisation levels [39] is presented in

Table 3. The classification of soil salinisation degree.

Soil Classification	Non-Salinised	Slightly Salinised	Moderately Salinised	Severely Salinised	Extremely Severely Salinised
Salt Content (g/kg)	<1	1–3	3–5	5–10	>10
Saturated Extract EC25 (dS/m)	0–2	2–4	4–8	8–16	>16

.

2.3. Data Acquisition

GD and TDS data were collected from 30 measurement stations and analysed spatially using Kriging interpolation to produce spatial distribution maps with a resolution of 30 m × 30 m. These maps were utilised for point value extraction. The digital elevation model (DEM) data were obtained from the Geospatial Data Cloud (http://www.gscloud.cn) with a spatial resolution of 30 m by 30 m. The normalised difference vegetation index (NDVI) was obtained from the Landsat 8 satellite (LANDSAT/LC08/C02/T1_L2) with a spatial resolution of 30 m × 30 m. SM data were obtained from the National Tibetan Plateau Data Centre (http://www.tpdc.ac.cn) with a spatial resolution of 250 m × 250 m. Potential evapotranspiration (PET) data were acquired from the MODIS MOD13A2 product, including a spatial resolution of 500 m × 500 m. To maintain uniform spatial resolution across all datasets, the SM and PET data were resampled utilising the nearest-neighbour technique. This resampling adjusted the lower-resolution data to conform to the 30 m × 30 m resolution, minimising scale effects and maintaining data consistency in the study.

2.4. Research Methodology

This study employs ordinary least squares (OLS) regression and GWR models to thoroughly examine the spatial heterogeneity of the data and the spatial impact of independent variables on the dependent variable, conducting modelling and validation from both global and local viewpoints. The precise research procedures are as follows:

(1) Data Preprocessing and Spatial Weight Matrix Construction

Initially, the study data undergoes preprocessing, encompassing missing value imputation and outlier elimination, to guarantee data quality and the precision of the model estimate. To ensure the stability of the regression model, both the dependent and independent variables are standardised. The spatial weight matrix is a crucial instrument for delineating geographical relationships, and its definition directly influences the assessment of spatial correlation and the efficacy of the spatial model fitting. In this research, based on the spatial adjacency relationships between spatial units, two types of spatial weight matrices are constructed: one is the adjacency matrix W_ij based on threshold distance, and the other is the kernel weight matrix w_ij based on the Gaussian kernel function. The calculation formula is as follows:

$$W_{ij}=\left\{\begin{array}{c}1,{ d}_{ij}\le { d}_{Threshold }\\ 0, { d}_{ij}>{ d}_{Threshold }\end{array}\right.$$

(1)

$${\omega }_{ij}=\mathit{exp}(-{\displaystyle \frac{d_{ij}^2}{2h^2}})$$

(2)

where d_ij represents the spatial distance between locations i and j, d_Threshold is the distance threshold, determined through spatial neighbourhood analysis, and h is the bandwidth parameter, which controls the decay range of the kernel weight. By accounting for the gradual attenuation properties of spatial distance, the Gaussian kernel weight matrix can more precisely represent the local relationships among variables, rendering it especially appropriate for depicting the spatial heterogeneity of areas with significant spatial variation.

(2) Spatial Autocorrelation Test

Spatial autocorrelation denotes the resemblance of observed values among spatial units, indicating that neighbouring or physically proximate observation locations often display specific patterns or correlations. Neglecting spatial autocorrelation in regression analysis might result in biassed coefficient estimates and erroneous significance tests. Moran’s I index is a widely utilised metric for spatial autocorrelation intended to measure the global spatial autocorrelation of geographical data. Before selecting the regression model, assessing the extent of spatial autocorrelation in the data is essential. The calculation formula is as follows:

$$I={\displaystyle \frac{N}{W}}\cdot {\displaystyle \frac{{\sum }_{i=1}^N{\sum }_{j=1}^N{W}_{ij}(y_i-\overline{y})(y_j-\overline{y})}{{\sum }_{i=1}^N(y_i-\overline{y}{)}^2}}$$

(3)

where N is the number of sample points, $y_i$ and $y_j$ are the observed values at locations i and j, respectively, $\overline{y}$ is the mean value of the dependent variable, and W is the sum of all weights.

However, calculating Moran’s I alone does not directly assess the statistical significance of spatial autocorrelation. Therefore, it is necessary to standardise Moran’s I by transforming it into Z(I) for significance testing. The formula for this transformation is as follows:

$$Z\left(I\right)={\displaystyle \frac{I-E\left(I\right)}{\sqrt{Var\left(I\right)}}}$$

(4)

where E(I) is the theoretical mathematical expectation, and Var(I) is the variance of Moran’s I.

The Z(I) value can determine the significance of spatial autocorrelation. When Z(I) = 0, the Z(I) value equals its mathematical expectation, indicating that the observed values follow an independent random distribution. The Moran scatter plot indicates that the observed values are uniformly distributed over the four quadrants, implying a lack of significant spatial autocorrelation. When Z(I) > 0 is statistically significant, it signifies positive spatial autocorrelation, with observed values clustered in the 1st and 3rd quadrants of the Moran scatter plot, reflecting high-high or low-low clustering in the LISA map. When Z(I) is significant, it signifies negative spatial autocorrelation, with observed values clustered in the 2nd and 4th quadrants of the Moran scatter plot, reflecting low-high or high-low clustering in the LISA map.

(3) Detection of Multicollinearity

Tolerance and variance inflation factor (VIF) are commonly used indicators for detecting multicollinearity in regression analysis [40]. Multicollinearity refers to a high correlation between independent variables, which can lead to unstable regression coefficients, thus affecting the model’s predictive accuracy and interpretability.

Tolerance: Tolerance represents the inverse of the linear relationship between a given independent variable and the others, calculated as 1−R² (coefficient of determination). A small tolerance value (typically less than 0.1) indicates strong multicollinearity, while a larger tolerance value suggests lighter multicollinearity.

VIF: VIF is the reciprocal of tolerance and measures the impact of correlations between an independent variable and the others. A large VIF (typically greater than 10) indicates severe multicollinearity, which may affect the stability of the regression model. If the VIF is less than 7.5, it suggests that the impact of multicollinearity on the model is minimal and does not typically affect the analysis results.

(4) Global Model Construction

OLS is a traditional regression analysis technique to estimate the overall connection between independent and dependent variables. OLS presumes that the regression coefficients are uniform over the whole geographical domain and that the error terms are independent and identically distributed. The conventional model for OLS regression is as follows:

$$Y_i={\beta }_0+{\beta }_1{X}_{i1}+{\beta }_2{X}_{i2}+\cdots +{\beta }_k{X}_{ik}+{\epsilon }_i$$

(5)

where Y_i is the dependent variable, X_i1, X_i2,…, X_ik are the independent variables, β β₀, β₁, β₂,…, β_k are the regression coefficients, and ε_i is the error term. OLS estimates the regression coefficients by minimising the residual sum of squares.

Nonetheless, OLS presupposes the independence of error terms; hence, spatial autocorrelation in the data may lead to biassed estimations of the regression coefficients. In spatial autocorrelation, OLS may inadequately represent the genuine connections among variables, constraining its use in spatial data analysis.

(5) Local Model Construction

The basic assumption of the GWR model is that the regression coefficients vary across different spatial locations, and the coefficients are estimated by weighting local data. The expression for the GWR model is as follows:

$$Y_i={\beta }_0\left(s_i\right)+{\beta }_1\left(s_i\right)X_{i1}+{\beta }_2\left(s_i\right)X_{i2}+\cdots +{\beta }_k\left(s_i\right)X_{ik}+{\epsilon }_i$$

(6)

where ${\beta }_k\left(s_i\right)$ is the local regression coefficient at location $s_i$, estimated using Weighted Least Squares (WLS). The estimate of the regression coefficients includes data from adjacent sites, with spatial weights determined by a Gaussian kernel function. By accounting for the local features of each place, the GWR model can more precisely delineate the connections between variables in space, especially in regions exhibiting significant spatial variety.

(6) Model Comparison and Selection

This study uses the Akaike Information Criterion (AIC) to evaluate the OLS and GWR models’ goodness of fit to identify the ideal regression model. The AIC is a metric that evaluates both the model’s quality of fit and its complexity. A reduced AIC value often signifies superior model fit while mitigating overfitting. The formula for computing the AIC is as follows:

$$\mathrm{A}\mathrm{I}\mathrm{C}=2n\ln \widehat{\delta }+n\ln \left(2\mathsf{\pi }\right)+n\left[{\displaystyle \frac{n+tr\left(S\right)}{n-2-tr\left(S\right)}}\right]$$

(7)

$$\widehat{\delta }={\displaystyle \frac{\mathrm{R}\mathrm{S}\mathrm{S}}{n-tr\left(S\right)}}$$

(8)

where $\widehat{\delta }$ represents the maximum likelihood estimate of the residual variance, and the residual sum of squares (RSS) represents the model’s fitting error. n is the sample size, and tr(S) is the trace of the hat matrix S, which reflects the model’s degrees of freedom and is related to the complexity of the kernel function.

A much lower AIC value for the GWR model compared to the OLS model signifies that the GWR model offers a superior fit to the data, effectively capturing the geographic heterogeneity and autocorrelation effects inherent in the dataset.

Data analysis and processing were conducted with SPSS 22.0 and Origin 2022. The spatial autocorrelation study of soil salinity was performed using Geoda 1.6.5 software, and the GWR model was developed using the GWR4.0 tool.

3. Results

3.1. Soil Salinisation Distribution Characteristics and Spatial Autocorrelation Analysis

Figure 2a illustrates that soil salinisation in the Donghaixin irrigation basin diminishes from north to south. The level of soil salinisation in the surface water irrigation zone exceeds that in the groundwater irrigation zone. The distribution of soil salinisation has clustering traits, with small clusters primarily located in the western region of the surface water irrigation zone and at the interface between surface and groundwater irrigation regions. Figure 2b illustrates that just 1.3% of the land inside the Donghaixin irrigation basin is devoid of soil salinisation. The land is predominantly impacted by light, moderate, and severe salinisation, with proportions of 83.8%, 13.8%, and 1.3%, respectively. In the surface water irrigation region, the percentage of moderately salinised soil is 23.1%, markedly above the 4.9% recorded in the groundwater irrigation region. Conversely, almost 90% of the soil in the groundwater irrigation zone exhibits mild salinisation. Figure 2c illustrates that, under mild salinisation conditions, the average soil salinity in the surface water irrigation zone is 2.45 g/kg, but it is 1.72 g/kg in the groundwater irrigation zone. The groundwater irrigation zone contains fewer non-salinised and moderately salinised sample points, whereas the surface water irrigation zone has a reduced number of severely salinised sample points. Consequently, no comparison analysis is conducted due to possible bias in the methods.

Figure 3a illustrates that Moran’s I for soil salinity is 0.67, accompanied by a Z-value of 10.26. The Z-value substantially exceeds the standard significance level (usually 1.96), signifying that the soil salinisation in the study area has considerable positive spatial autocorrelation. Moran’s I indicates that the salinisation traits of neighbouring areas exhibit similarity, demonstrating a spatial clustering effect. Data points are predominantly clustered in the first and third quadrants, indicating that neighbouring regions exhibit comparable salinisation levels, perhaps affected by analogous natural or human processes. Figure 3b illustrates that the red zone northwest of the study area signifies a high-high cluster, suggesting that these places and their adjacent regions exhibit markedly elevated salinisation levels. The blue zone in the southern section of the study area denotes a low-low cluster, signifying that these areas and their adjacent regions exhibit reduced salinisation levels. The extensive white region in the middle area has negligible spatial autocorrelation, indicating that the salinisation properties in these zones are predominantly random and lack a distinct clustering pattern.

3.2. Correlation Analysis Between Soil Salinity and Micro-Ion Content

Table 4. Soil ion content at different salinisation levels.

Degree	K+	Na+	Ca2+	Mg2+	Cl−	SO42−	HCO3− + CO32−
	mg/kg	mg/kg	mg/kg	mg/kg	mg/kg	mg/kg	mg/kg
Non-salinised	10.16	103.24	92.84	44.27	198.20	126.15	138.47
Slightly salinised	25.81	269.48	130.57	101.56	500.05	304.09	503.05
Moderately salinised	48.82	658.69	201.69	156.41	962.29	655.07	802.42
Severely salinised	81.82	1132.39	330.11	289.23	1490.09	981.10	890.72

illustrates that with the escalation of soil salinisation, the concentration of different ions progressively rises. In soils exhibiting non-severe salinisation, the concentrations of Cl⁻ and HCO₃⁻ + CO₃²⁻ surpass those of other ions. In non-salinised soil, the concentrations of Na⁺ and Cl⁻ are 103.24 mg/kg and 198.20 mg/kg, respectively. Still, in substantially salinised soil, the concentrations of Na⁺ and Cl⁻ escalate to 1132.39 mg/kg and 1490.09 mg/kg, reflecting a tenfold and sevenfold rise, respectively. The fast buildup of Na⁺ and Cl⁻ signifies that sodium and chloride salts are pivotal in severe salinisation due to their high solubility and mobility. The elevation of SO₄²⁻ levels is notable, particularly in highly salinised soils, underscoring the critical function of sulphate in the salinisation process. The rise in HCO₃⁻ + CO₃²⁻ is comparatively modest, perhaps linked to the precipitation equilibrium of carbonates, and plays a substantial role in soil alkalisation. The sequence of ion content variation with escalating soil salinity is as follows: Na⁺ > K⁺ > SO₄²⁻ > Cl⁻ > Mg²⁺ > HCO₃⁻ + CO₃²⁻ > Ca²⁺.

Figure 4a illustrates the study area’s substantial positive association between soil salinity and many ions (K⁺, Na⁺, Ca²⁺, Mg²⁺, Cl⁻, SO₄²⁻, and HCO₃⁻ + CO₃²⁻). The correlation coefficients for Na⁺, Cl⁻, and SO₄²⁻ are 0.84, 0.85, and 0.81, respectively. The correlations of K⁺, Mg²⁺, and HCO₃⁻ + CO₃²⁻ with salinity are 0.70, 0.68, and 0.67, respectively, signifying moderate relationships. The link between Ca²⁺ and salinity is generally weak, affected by more intricate soil chemical processes. Furthermore, the correlation value between Na⁺ and Cl⁻ is 0.63, while that between Cl⁻ and SO₄²⁻ is 0.66, indicating a pattern of synergistic fluctuation among the ions, potentially implying a shared source or interaction mechanism.

Figure 4b illustrates that the fluctuation in the association between ions and soil salinity indicates ion sensitivity. The soil data are categorised into two categories to examine the mechanisms of soil salinisation in the Donghaixin irrigation region. In soils exhibiting moderate salinisation, a strong link exists between soil salinity and the ions Cl⁻, SO₄²⁻, Mg²⁺, and Na⁺, signifying their substantial influence on soil salinisation at this phase. The link between K⁺ and soil salinity is comparatively low, whereas Ca²⁺ exhibits a non-significant negative correlation, indicating a little impact of these two ions during the moderate salinisation phase. Na⁺ is the ion most strongly correlated with soil salinity in moderately salinised soils, signifying its predominant involvement at this stage. Simultaneously, Cl⁻ and SO₄²⁻ exhibit strong connections, continuing to fulfil a significant function. The correlation between HCO₃⁻ and CO₃²⁻ markedly rises, indicating heightened sensitivity under mild salinisation conditions. Conversely, the connection of K⁺ diminishes, although the correlations of Ca²⁺ and Mg²⁺, however modest, continue to affect soil salinity to a degree. The primary influencing variables in weakly salinised soils are Cl⁻, SO₄²⁻, Mg²⁺, and Na⁺. With the escalation of soil salinisation, Na⁺, SO₄²⁻, HCO₃⁻ + CO₃²⁻, and Cl⁻ emerge as the principal determinants of salinisation. The relationship between K⁺ and Mg²⁺ diminishes with heightened salinisation, indicating that these ions may exert a more significant influence during the first phases of salinisation. Figure 4c illustrates the hierarchy of ion sensitivity to soil salinisation as follows: Ca²⁺ > Na⁺ > HCO₃⁻ + CO₃²⁻ > Mg²⁺ > K⁺ > Cl⁻ > SO₄²⁻.

3.3. Regression Analysis and Spatial Prediction of Soil Salinisation

The buildup of surface soil salinity results from the interplay of numerous mechanisms related to the water-salt movement process. This study chose eight components strongly associated with soil salinity for regression analysis: OM, pH, GD, TDS, DEM, NDVI, SM, and PET. Tolerance and VIF assessments were performed to mitigate multicollinearity concerns among the explanatory variables. The tolerance and VIF values for each component varied from 0.41 to 0.82 and 1.08 to 2.44, respectively, with tolerance values over 0.1 and VIF values below 7.5, signifying the absence of significant multicollinearity or only minimal multicollinearity among these variables.

Soil salinity in the Donghai Xin Irrigation Area served as the dependent variable. At the same time, those eight environmental components were utilised as independent variables to develop both OLS and GWR models. The OLS model findings demonstrated a positive correlation between soil salinity and OM, TDS, and the NDVI while exhibiting a negative correlation with pH, GD, DEM, SM, and PET, with GD and TDS showing a significant correlation (p < 0.05).

Table 5. Comparison of OLS and GWR model statistics for soil salinity.

Model	AIC	R2	Adjusted R2	RSS	df	F
OLS model	150.26	0.61	0.56	23.87	71.00	13.85
GWR model	126.90	0.78	0.68	13.67	54.83	2.67

Note: df: degrees of freedom; F: F-statistic.

illustrates that the AIC value of the GWR model markedly decreased from 150.26 for OLS to 126.90, signifying a substantial enhancement in the model’s fit. The corrected R² for the GWR model rose to 0.68, reflecting a 12% enhancement compared to the OLS model’s 0.56, indicating a superior capacity to elucidate the spatial variability of soil salinity. Moreover, the RSS for the GWR model diminished from 23.87 to 13.67, while the degrees of freedom reduced from 71 to 54.83, substantiating the substantial enhancement in data fitting accuracy afforded by the GWR model.

Figure 5 illustrates that the forecast values from the GWR model align more closely with the observed values than those from the OLS model at most sample locations. Particularly in areas with elevated soil salinity levels, the GWR model demonstrated a greater accuracy in representing the variations in the observed values. This suggests that the GWR model is more adept at capturing the geographical variability of soil salinity. Compared to the OLS model, the GWR model demonstrated a notable enhancement in predictive accuracy and application, reinforcing the superiority of GWR in tackling spatial heterogeneity challenges.

Figure 6 illustrates that the influence and intensity of many environmental influences on soil salinity demonstrate considerable geographical variation. The impact of these factors on soil salinity is intricately linked to regional topography, hydrological conditions, and plant cover. The regression coefficient of DEM has a predominantly negative association, which diminishes from northwest to southeast. The regression coefficient of the NDVI indicates a negative association in the northwest and southeast and a positive correlation in the northeast-southwest zone. The regression coefficient of GD is primarily negatively correlated, with a positive association seen just in the northeastern area. The regression coefficient of TDS has a largely positive association, which diminishes from northwest to southeast. The regression coefficient of PET exhibits a positive association just in a limited region in the southwest, with the correlation intensifying from south to north. The regression coefficient of OM has a largely positive connection, which diminishes from northwest to southeast. The regression coefficient of pH has a primarily negative correlation, characterised by a striped distribution from west to east. The regression coefficient of SM is primarily negatively correlated, exhibiting a positive association just in the northern section of the area.

Figure 7 illustrates a strong correlation between the projected and observed data for geographic distribution patterns, but some localised discrepancies persist. The observed and forecasted salt distribution indicates elevated concentrations in the western and central areas, whereas diminished levels are predominantly found in the eastern and southern parts. This suggests that the predictive model may accurately represent the whole geographical distribution pattern of soil salinisation. In the highly salinised portions of the northwest, the projected values somewhat exceed the actual values, likely attributable to the model’s overfitting of local characteristics in high-salinity locations. In the central and eastern regions, the projected values closely align with the observed values; nevertheless, the model significantly underestimates the actual salinity levels in some low-salinity zones. The forecast map displays a more uniform distribution trend, but the measured map reveals discontinuities resulting from the constraints of sample sites. This indicates that the model mitigates the geographical distribution deficiencies resulting from inadequate data, offering a more thorough viewpoint for regional salinisation evaluation.

3.4. Analysis of Groundwater and Soil Water-Soluble Ion Characteristics

Figure 8a,b illustrate the Piper trilinear diagram, which effectively delineates the similarities and differences in the distribution and development of groundwater and soil water-soluble ions. Na⁺, Cl⁻, SO₄²⁻, and HCO₃⁻ are the principal ions in both irrigation types; nevertheless, the predominant anions in the soil and groundwater, as well as among other irrigation locations, exhibit unique features. The groundwater chemistry in the surface water irrigation zone predominantly consists of Na-SO₄·HCO₃, with SO₄²⁻ and HCO₃⁻ representing a significant percentage, indicating the impact of surface water recharge and external sulphate contributions. The water chemistry in the groundwater irrigation area is predominantly Na-HCO₃·Cl, characterised by significant enrichment of bicarbonate and chloride ions. This suggests that evaporation concentration and ion exchange processes have markedly affected groundwater irrigation, resulting in heightened salt accumulation and sodification.

Na predominantly influences the distribution of cations in soil water. The surface water irrigation region has a comparatively elevated concentration of Mg²⁺, characterised by a water chemistry type predominantly of Na·Mg-Cl·SO₄. In contrast, the groundwater irrigation area demonstrates a greater concentration of Ca²⁺, Mg²⁺, Cl⁻, and HCO₃⁻, resulting in a Na·Ca·Mg-Cl·HCO₃ water chemistry profile. This indicates that groundwater irrigation presents a greater risk of salinisation and alkalisation. The surface water irrigation area exhibits a comparatively elevated SO₄²⁻ concentration, significantly affected by the input from water sources and external environmental factors. In contrast, the groundwater irrigation area substantially enriches Cl⁻ and HCO₃⁻, primarily influenced by evaporation concentration and ion exchange processes.

Figure 9 illustrates that when salinisation intensifies, the concentration of each ion exhibits an increasing trajectory. The concentrations of Cl⁻ and Na⁺ markedly rise, peaking at extreme salinisation conditions, signifying that Na-Cl type salts predominate in the soil salinisation process. The concentrations of CO₃²⁻, HCO₃⁻, and SO₄²⁻ progressively rise with the escalation of salinisation, indicating the enrichment of alkaline salts and sulphate ions. The concentrations of Ca²⁺ and Mg²⁺ remain comparatively low, exhibiting a steady rise, indicating their contribution in the initial phases of salinisation, but are progressively supplanted by Na⁺ in subsequent stages. The process of soil salinisation is chiefly caused by the buildup of Na⁺ and Cl⁻, alongside a persistent rise in HCO₃⁻ and SO₄²⁻, demonstrating apparent traits of salinisation and alkalisation.

4. Discussion

4.1. Spatial Heterogeneity of Soil Salinity and Environmental Factors and Their Driving Mechanisms

The development and progression of soil salinisation are affected by environmental and anthropogenic variables, with soil texture, terrain, groundwater, climate, and vegetation being the principal environmental determinants [41]. This study used the OLS and GWR models to examine the correlation between soil salinity and several regional environmental parameters. The impact of several environmental variables on soil salinity has considerable regional variability, illustrating the processes by which these factors lead to the buildup or reduction in soil salinity. Soil salinity correlates positively with OM, TDS, and the NDVI while negatively correlating with pH, GD, DEM, SM, and PET. GD and TDS significantly influence soil salinity. This study reveals that the average pH of the study area is 8.47, signifying alkaline soil. Sodium salts liberate Na⁺ via ion exchange, perhaps substituting H⁺ in the soil and decreasing the pH, akin to the findings of Zhu et al. [42]. Nonetheless, the alkaline properties of sodium salts and the soil’s buffering capacity may result in a gradual and restricted impact of salts on pH, yielding an inconsequential association between the two. The processes of soil formation and hydrological circumstances somewhat obscure the direct influence of salt on pH variations. Li et al. [43] discovered that for each 10 cm rise in GD, SM and salt content diminish by 0.56% and 0.06 g/kg, respectively. This study obtained the same outcome, indicating that the regional distribution of regression coefficients for GD and SM was analogous. GD directly influences soil water content, and salt deposition results from groundwater evaporation [44]. When the GD is superficial, capillary action transfers water from the saturated zone to the unsaturated zone, facilitating the passage of salts to the root zone and accumulating on the soil surface following evaporation. As DEM grows, drainage conditions often enhance, allowing salts to be efficiently leached by precipitation infiltration and surface runoff, reducing soil salinity. Bakr and Ali [45] identified a significant negative association between DEM and soil electrical conductivity, along with the elevated soil salinity noted in the lower elevation regions in the northwest within this study area. Weather circumstances predominantly affect PET; GD can indirectly alter PET by influencing soil water availability. Elevated PET generally coincides with reduced GD. In the northwest of the study area, shallow GD is associated with elevated soil salinity regression coefficients; however, this region also displays soil salt crusting [46], which may enhance soil albedo and inhibit soil PET. Consequently, the regression coefficients in this area may be undervalued. The interplay between soil OM and salinity is intricate, encompassing SM retention capacity, ion exchange, soil structure, and microbial activity [47]. The regression coefficients are comparatively elevated northwest of the study area, partly due to years of Yellow River irrigation enriching the OM. The low-lying topography with shallow groundwater may diminish microbial activity, thereby impeding the decomposition of OM and resulting in its accumulation. The DEM and GD in the study area are shown in Figure 10. Zhou et al. [48] discovered that elevated TDS in groundwater results in increased soil salinity buildup, corroborating the findings observed in surface water irrigation regions in our investigation. Moreover, groundwater at deeper depths has elevated TDS levels, and salts may ascend to the soil surface by capillary action, exacerbating soil salinity. While most research indicates that high-salinity soils impair SM retention, hinder plant development, and diminish the NDVI [49], the conclusions of this study diverge. This may be attributed to the comparatively salt-tolerant crops cultivated in the study area, such as sunflower and maize, which exhibit lesser sensitivity to salinity. The study area employs winter irrigation and spring leaching techniques, establishing optimal water and soil conditions throughout the crop development, complicating the link between the NDVI and soil salinity. The variation in soil salinity in the study area is often driven by hydrodynamic and drainage variables (negative correlation) and accumulation and salt retention factors (positive correlation). The buildup or dilution of soil salts is contingent upon two primary processes: TDS and local climatic conditions, which dictate salt intake, whereas drainage conditions and evaporation govern salt transport.

The GWR model simulation of soil salinity distribution indicates that the model accurately represents variations in soil salinity across the study area, with salinity progressively rising from south to north. The elevated salinity zones in the northwest and other localised high-salinity areas correspond with the observed data. Nevertheless, the model tends to underestimate salinity in certain regions, especially in the high-salinity zones in the northwest and certain fishpond areas. This variance may be associated with land use, irrigation techniques, and past environmental attributes [50]. The study area encompasses both surface water irrigation and groundwater irrigation zones. In the surface water irrigation zone, yearly winter irrigation is performed to dilute surface salts and facilitate the movement of salts to deeper layers, therefore alleviating soil salinisation. In low-lying regions with inadequate drainage systems, winter irrigation water often evaporates, resulting in the reemergence of salts near the surface and subsequent secondary buildup [51]. This effect is most pronounced in the northwest region of the study area. In the groundwater irrigation zone, the absence of winter irrigation results in soil salinity being predominantly affected by natural precipitation and groundwater dynamics, leading to significantly lower and more stable salt deposition. Furthermore, the smoothing property of the GWR model constrains its capacity to precisely replicate regional variations and local extremes, leading to inconsistencies between projected and observed data. Although we conducted a model accuracy assessment, independent validation was not performed, primarily due to the small size of the dataset and the difference between the research objectives and the requirements for validation analysis. The primary goal of this study was to explore the relationship between soil salinity distribution and environmental factors and to predict spatial distribution, with a focus on analysing spatial variability rather than independently validating the model’s accuracy. We fully acknowledge that validation analysis is crucial to ensuring the reliability and accuracy of the predicted results. Therefore, future research should collect more experimental data, include more comprehensive irrigation models, land use data, and other variables, conduct independent validation analysis, further improve the model’s predictive capability, and ensure its applicability under different environmental conditions. Additionally, differentiated management measures should be developed for different irrigation zones. In surface water irrigation zones, drainage systems must be enhanced to mitigate the possibility of secondary salt buildup in low-lying regions. In groundwater irrigation areas, it is essential to improve water quality monitoring by taking precipitation patterns into account to guarantee optimal salt content and preserve soil health. Environmental variables govern the regional distribution of salinisation in the southern Yellow River irrigation zone, whereas human causes dictate its spatial fluctuation.

4.2. The Relationship Between Salinisation and Groundwater

Salinisation and its association with groundwater constitute a significant study focus in environmental studies, particularly in agricultural irrigation zones within arid and semi-arid regions [52]. The correlation between water and soil salinity is intricate. In regions with shallow groundwater, water can ascend by capillary action, transporting salts to the soil surface. Likewise, due to precipitation and the penetration of irrigation water, soil salts may percolate downward into the groundwater. Agricultural irrigation in arid and semi-arid regions substantially affects groundwater. Research conducted by Qian and Li [53] demonstrates that channel leakage and field irrigation infiltration account for up to 80% of groundwater recharge in the Yinchuan Plain. This study identifies the principal sources of groundwater recharge in the study area: precipitation, channel leakage, field irrigation, and water from the Yellow River. The interplay of these sources collectively dictates groundwater salinity distribution and establishes a dynamic link between salt buildup and groundwater recharge. Drainage ditches are employed to regulate the shallow GD, hence mitigating salinisation and alkalisation. Simultaneously, ions like HCO₃⁻, Na⁺, Cl⁻, and SO₄²⁻ present in groundwater and soil will ultimately be transported into the Yellow River via hydrological processes. Consequently, the Yellow River irrigation zone on the southern bank of the Yellow River in Inner Mongolia is distinctly threatened by water contamination owing to soil salinisation.

Groundwater evaporation significantly contributes to soil salinisation in arid and semi-arid regions. Wang et al. [22] indicated that due to irrigation, the shallow depth of groundwater makes it prone to evaporation, which may accelerate the depletion of groundwater both quantitatively and qualitatively, enhance the upward migration of salts from groundwater and the subsurface unsaturated zone to the surface soil, and induce soil salinisation. In the study area along the southern bank of the Yellow River in Inner Mongolia, soil salinisation is primarily characterised by Na⁺, Cl⁻, SO₄²⁻, and HCO₃⁻, with its intricate development. It is intricately linked to groundwater evaporation and is also affected by factors such as the quality of irrigation water and the application of pesticides and fertilisers. In surface water irrigation regions, the shallow groundwater table and elevated ion concentration in the irrigation water result in a pronounced evaporation concentration effect, causing the buildup of soluble and mobile ions such as Na⁺ and Cl⁻ in the soil. Simultaneously, SO₄²⁻ quickly interacts with Na⁺ or Ca²⁺ to produce sulphate salts, intensifying salt buildup and soil structure degradation [54]. HCO₃⁻ readily reacts with Na⁺ to produce Na₂CO₃, hence elevating soil alkalinity (increasing pH), which may adversely impact plant development if alkalinity reaches excessive levels [55]. In regions utilising groundwater irrigation, where the GD is substantial, the principal source of surface soil salinity is from irrigation water and the accumulation of salts resulting from the evaporation of SM post-irrigation. Moreover, prolonged and excessive use of fertilisers and pesticides can introduce soluble ions such as Cl⁻ and SO₄²⁻, altering the soil’s physical and chemical characteristics and exacerbating salt buildup, intensifying the salinisation problem. With the escalation of salinisation, the concentrations of Na⁺, Cl⁻, and SO₄²⁻ in the soil markedly increase, signifying that these ions are the principal accumulators in the salinisation process. In contrast, the alterations in CO₃²⁻ + HCO₃⁻ are more stable, maybe attributable to their diminished migratory capacity in the soil. Liu [56] discovered that SO₄²⁻ and Cl⁻ have comparable migration rates attributable to evaporation, but evaporation exerts a comparatively little influence on HCO₃⁻ in the soil. This study indicates that the rates of increase in these two ions with rising salinisation are comparable, implying that evaporation is a significant factor in the soil’s buildup of SO₄²⁻ and Cl⁻. In the study area, the concentrations of HCO₃⁻ in groundwater and soil water-soluble ions are comparatively elevated, probably attributable to external sources such as Yellow River water. The principal sources of soil salinisation in the study area are evaporation concentration, ion movement, and agricultural practises, with groundwater evaporation exerting a considerable influence in regions with shallow groundwater.

4.3. Comprehensive Control Strategies for Soil Salinisation Management

Controlling soil salinisation in the irrigation zone on the southern bank of the Yellow River in Inner Mongolia is a multifaceted, integrated process that necessitates meticulous GD management, synchronised irrigation and drainage, salt leaching during non-growing seasons, optimisation of crop planting structures, and ecological restoration. Accurate management of GD is essential for attaining water-salt equilibrium. Effectively regulating the groundwater table’s depth can significantly diminish salt migration induced by capillary action [57]. The drainage canal system’s architecture must be optimised according to the irrigation region’s varying soil qualities and GD thresholds. Various methods, including open ditches, subterranean pipelines, and vertical wells, can guarantee adequate drainage during irrigation, fall irrigation, and periods of intense rainfall. Furthermore, certain agricultural ditches may become obstructed owing to sediment accumulation and insufficient excavation, necessitating routine desilting. Coordinated irrigation and drainage can be accomplished with a comprehensive system integrating surface water and groundwater resources. Enhancing the water inflow and outflow mechanisms via control structures at the canal’s entrance and utilising diverse water sources and water-conserving irrigation technology reduces the danger of salt buildup caused by irrigation and enhances water usage efficiency [58]. Furthermore, integrating free-flowing and forced drainage for primary drains facilitates dynamic management and resolves efficiency concerns associated with long-distance drainage and numerous drainage outlets. Implementing strategies such as extracting groundwater before the rainy season and enhancing the upkeep of drainage canals during the rainy season can effectively regulate GD and mitigate the dangers of salt retention. Regulating salt levels during the dormant season is essential for controlling soil salinity. Moderate irrigation in winter and spring may substantially decrease salt formation on the soil surface and preserve SM, establishing optimal soil conditions for crop growth throughout the growing season [59]. Ultimately, optimising agricultural planting structures and implementing ecological restoration are fundamental techniques for enhancing the utilisation efficiency of saline-alkali land [60]. In regions with light to moderate salinisation, salt-tolerant crops, including sunflowers, silage corn, and sorghum, can be cultivated to optimise land resources. Salt can be leached into deeper soil strata via rice field irrigation in low-lying saline-alkali regions. Cultivating halophytic species like seepweed and reeds in highly salinised regions might facilitate ecological restoration. The judicious use of diverse technology systems can mitigate soil salinisation hazards and offer management measures for water-salt equilibrium and sustainable agricultural advancement in the southern Yellow River irrigation region of Inner Mongolia.

5. Conclusions

Soil salinisation in the East China irrigation region diminishes progressively from north to south, with more than 80% of the land exhibiting mild salinisation. The salinisation level of soil in surface water irrigation regions surpasses that of groundwater irrigation areas, and salinisation has notable spatial aggregation features. Ion concentration escalates with soil salinity, and the rate of change in ion concentration is prioritised as follows: Na⁺ > K⁺ > SO₄²⁻ > Cl⁻ > Mg²⁺ > HCO₃⁻ + CO₃²⁻ > Ca²⁺. The susceptibility of ions to soil salinisation is ordered as follows: Ca²⁺ > Na⁺ > HCO₃⁻ + CO₃²⁻ > Mg²⁺ > K⁺ > Cl⁻ > SO₄²⁻. With the escalation of soil salinisation, Na⁺, SO₄²⁻, HCO₃⁻ + CO₃²⁻, and Cl⁻ emerge as the principal determinants of salinisation. The GWR model markedly enhances the fitting of soil salinity in the irrigated area on the southern bank of the Yellow River compared to the OLS model, particularly in capturing the spatial variability of salinity. The GWR model can more precisely represent the trend of observed data, especially in regions with elevated soil salinity. In some low-salinity areas, the model marginally underestimates the true extent of salinisation. Inner Mongolia’s southern Yellow River irrigation region encounters considerable water contamination hazards. Evaporation-induced groundwater migration is not the sole catalyst for soil salinisation in the irrigation zone. Groundwater and soil water chemistry indicate Na⁺ as the predominant cation, yet notable disparities exist in the distribution of anions. In surface water irrigation regions, SO₄²⁻ concentrations are comparatively elevated, primarily affected by water supplies and agricultural practises. In groundwater irrigation regions, Cl⁻ and HCO₃⁻ are highly concentrated, considerably influenced by evaporation concentration and ion exchange mechanisms.