Soil Physicochemical Improvement in Coastal Saline–Alkali Lands Through Salix matsudana × alba Plantation

Abstract

To evaluate the ecological remediation effect of Salix matsudana × alba on saline coastal soils, we established a five-year field experiment in Rudong County, Jiangsu Province, China. The experiment was designed with three salinity gradients (low, medium, and high) and five plant spacing treatments (2 × 2 m, 2 × 3 m, 3 × 3 m, 3 × 4 m, and 4 × 4 m). Soil samples were collected annually at a depth of 0–20 cm using grid and random sampling methods. Indicators of soil physicochemical properties and heavy metal content were measured, including soil organic matter (SOM), pH, total nitrogen (TN), total phosphorus (TP), total potassium (TK), electrical conductivity (EC), total salinity (TS), and bulk density (BD). Additionally, eight heavy metals were analyzed: zinc (Zn), chromium (Cr), nickel (Ni), copper (Cu), cadmium (Cd), lead (Pb), arsenic (As), and mercury (Hg). Results showed that the hybrid willow significantly improved SOM content by up to 90% and reduced EC and TS by 52% and 60% over five years, especially under low and medium salinity conditions with dense planting (2 × 2 m, 2 × 3 m). The content of most heavy metals exhibited a decreasing trend or remained stable, indicating the plant’s phytostabilization potential (i.e., stabilization of heavy metals via plant-soil interaction). Principal component analysis (PCA) and random forest (RF) modeling identified SOM, EC, TS, and BD as the dominant factors influencing soil quality improvement. A soil quality index (SQI) was constructed based on PCA-derived weights, which further confirmed the positive ecological effect of this hybrid species on coastal saline soils. This study provides scientific evidence supporting the use of Salix matsudana × alba as a promising species for large-scale ecological restoration in coastal saline-alkaline lands.

1. Introduction

As a frontier zone of land-sea interaction, coastal tidal flats face dual pressures from global climate change and intensifying human activities [1]. Against this backdrop, soil salinization has become one of the primary environmental challenges hindering sustainable development in coastal areas. Salinization threatens food security, causes substantial economic losses [2], and leads to ongoing land degradation. The formation of saline soils is primarily attributed to the accumulation of soluble salts (e.g., NaCl) originating from wind deposition, rainfall, rock weathering, and anthropogenic activities. When salt levels in the soil reach a certain threshold, they disrupt soil structure, impair hydraulic properties, and disturb the ionic balance, triggering cascading effects, including osmotic stress, ion imbalance, and reduced microbial activity. These soil processes interact within the rhizosphere, exerting significant stress on plants and ultimately affecting crop productivity [3]. In the coastal areas of the East Asian monsoon region, high salt stress is one of the main challenges facing vegetation restoration, severely limiting the success rate of vegetation establishment. Salt damage is also a key limiting factor affecting the distribution and growth of coastal vegetation in the subtropical monsoon climate along the southern coast of China [4]. Due to the widespread inhibitory effects of high salinity—such as difficulties in plant establishment, slow growth, and high mortality—it is challenging to implement large-scale vegetation restoration on severely salinized tidal flats. Although success rates may vary across regions due to environmental conditions and specific cultivation practices, the negative impacts of salinity on plant growth are widely observed. In China’s coastal regions, soil salinization is becoming increasingly severe, with some provinces experiencing a concentrated distribution of saline soils, among which Jiangsu faces particularly severe remediation pressure. In the coastal tidal flats of Rudong County, Nantong City, land reclamation and agricultural development have caused distinct salinity gradients and prominent ecological degradation, making these areas a key constraint to regional sustainable development.

Effective management and remediation of salinized soils is essential for achieving most of the United Nations Sustainable Development Goals [5]. As a vital approach to the ecological restoration of saline soils, artificial vegetation reconstruction improves soil environments through species-specific mechanisms. For example, Robinia pseudoacacia possesses nitrogen-fixing root nodules that enhance soil nitrogen content. Meanwhile, its root activity may also influence rhizosphere pH. However, R. pseudoacacia has relatively limited salt tolerance; studies have shown that a soil salinity level exceeding 0.3% is considered its salt tolerance threshold [6], above which its growth is significantly inhibited and survival rates drastically reduced, thus restricting its application in high salinity tidal flat regions. Kandelia obovata shows salt tolerance but is sensitive to waterlogging due to limited gene expression and weak physiological responses under flooding stress [7]. This leads to fluctuating ecological restoration outcomes in monsoon-zone intertidal areas. In recent years, species of Salix have been widely used in saline soil improvement due to their high salt tolerance, large biomass, and strong ecological adaptability. Willows can significantly improve soil structure and fertility by regulating salt distribution and promoting soil aggregate formation. Previous studies have indicated that particular willow species can tolerate moderate salinization (ECₑ ≤ 5.0 dS/m), making them suitable for the remediation of salinized farmlands [8]. Ran et al. further found that Salix matsudana exhibited stable growth under high salinity stress, and specific growth ratios could serve as key physiological indicators of its salt tolerance [9,10]. These findings provide valuable insights into the salt tolerance mechanisms of willows. However, the effects of different planting densities on these mechanisms and their significance for soil improvement remain to be investigated.

Hybrid willow trees, particularly the hybrids of Salix matsudana and Salix alba, exhibit stronger tolerance to environmental stress compared to their parental species. Studies by Quiñones Martorello et al. demonstrated that S. matsudana × alba hybrids can adapt effectively to vertically heterogeneous saline environments by adjusting root distribution [11]. Transcriptome studies show that Salix matsudana × alba regulates genes involved in cell wall remodeling and osmotic adjustment under combined salt and flooding stress, revealing strong molecular adaptability [12]. In addition to salt tolerance, willow species also possess potential for heavy metal remediation. Mleczek et al. found that S. alba and S. viminalis can effectively absorb heavy metals from soil [13]. Urošević et al. further confirmed that different willow genotypes vary significantly in their capacity to extract heavy metals, highlighting the genetic diversity advantages of willows in phytoremediation applications [14]. Moreover, different planting densities significantly affect the survival rate and biomass yield of willows, and scientifically optimized density configurations can enhance remediation outcomes. As a key management factor regulating plant growth and optimizing soil improvement efficiency, planting density plays a critical role in the overall performance of willow trees. Research by Kulig et al. indicated that planting density significantly influences the survival rate and biomass accumulation of willows and that appropriate density arrangements help optimize resource use efficiency and accelerate soil recovery [15].

As a pioneer species for ecological restoration of coastal saline-alkali lands, Salix matsudana × alba not only features rapid growth but also demonstrates strong resistance to high salinity, drought, and waterlogging stress and has demonstrated promising application potential in salinized areas of North America and Europe. However, its adaptation mechanisms under high salinity stress in the coastal zones of the East Asian monsoon region, as well as the long-term effects of different planting patterns on soil remediation, remain insufficiently studied. In particular, the long-term mechanisms underlying the interactive effects of multiple factors—such as salinity gradients and planting density—on changes in soil physicochemical properties and heavy metal accumulation dynamics are still unclear.

Building upon these insights, this study conducted a field experiment in the coastal tidal flats of Rudong County, Nantong City, Jiangsu Province. Three salinity gradients (0.1%–0.3%, 0.3%–0.5%, and 0.5%–0.8%) and five planting spacings (2 m × 2 m, 2 m × 3 m, 3 m × 3 m, 3 m × 4 m, and 4 m × 4 m) were established. Five years of continuous monitoring of Salix matsudana × alba plantations were carried out to systematically analyze the spatiotemporal dynamics of soil organic matter (SOM), total salinity (TS), nutrients, and heavy metal elements. This study hypothesizes that Salix matsudana × alba exerts significant effects on the improvement of soil physicochemical properties and heavy metal accumulation under different salinity gradients and spacing configurations; that the soil quality index (SQI) can effectively predict the most suitable cultivation type under varying salinity conditions; and that the predictive performance and rationality of SQI construction can be evaluated using the random forest (RF) model.

2. Materials and Methods

2.1. Study Site

This study was conducted in the coastal tidal flat area of Rudong County, Nantong City, Jiangsu Province, China (120°22′–121°42′E, 32°12′–32°36′N) (Figure 1). The study area is characterized by a subtropical monsoon climate, with an average annual temperature of 15.1 °C and an annual precipitation of approximately 1,050 mm. The experimental site is a typical reclaimed tidal flat that had been left fallow for two years after reclamation. Preliminary surveys showed that the surface soil (0–20 cm) exhibited a salinity range of 0.1%–0.8%, with evident spatial heterogeneity. Soil pH values mostly ranged between 8.0 and 9.5, indicating an alkaline environment. Soil organic matter (SOM) content was generally low, averaging 2.7 g/kg, reflecting poor soil fertility. Electrical conductivity (EC) fluctuated between 1.3 and 3.8 dS/m. Initial levels of total nitrogen (TN), total phosphorus (TP), and total potassium (TK) were low. The surface soils (0–20 cm) in the study area exhibit typical features of saline-alkaline soils characterized by high pH, salt accumulation, and low fertility. The soil texture ranges from sandy loam to silty loam, a composition commonly found in reclaimed coastal tidal flats, which significantly influences water retention, infiltration, and salt dynamics. Based on these morphological and physicochemical characteristics, the soils can be preliminarily classified as Solonchaks under the World Reference Base for Soil Resources (WRB 2022), with potential Anthrosol features resulting from long-term reclamation activities [16]. Heavy metal concentrations were within the range of natural background levels for coastal soils, although slight enrichment trends were detected in localized areas. The region is characterized by severe salt accumulation, poor physicochemical properties, and marked ecological degradation, exhibiting typical features of coastal saline soils and making it a suitable site for assessing ecological restoration and soil improvement effects in coastal tidal flats.

2.2. Experimental Materials and Design

Nanjing Forestry University provided two-year-old seedlings of Salix matsudana × alba. This hybrid was selected as the experimental species due to its strong salt and moisture tolerance as well as rapid growth characteristics. The seedlings had a stem diameter of 4.0–5.0 cm and were pruned to a height of 2 m. Each planting hole measured 60 cm × 60 cm. Transplanting was completed in October 2019, and the experiment was continuously monitored for five years, concluding in October 2024.

The experimental design employed soil salinity gradients and planting spacing as the primary treatment factors. Based on preliminary soil analysis, the experimental area was divided into three salinity levels: low salinity (0.1%–0.3%), medium salinity (0.3%–0.5%), and high salinity (0.5%–0.8%). Within each salinity level, five different planting spacings were implemented: 2 m × 2 m, 2 m × 3 m, 3 m × 3 m, 3 m × 4 m, and 4 m × 4 m. Each treatment was replicated three times, totaling 45 treatment units (3 salinity levels × 5 spacing treatments × 3 replicates). Each salinity experimental zone covered approximately 700 m² and included more than 40 trees, with buffer zones established to minimize interference between treatments. In addition to the treatment plots, corresponding non-planted control plots were established within each salinity zone. These control plots were monitored using the same sampling and analytical procedures over five consecutive years to serve as background references (Table S1).

The experimental site was located within a reclaimed coastal tidal flat area adjacent to several plots that had already been cultivated. Although the main experimental area was free from direct human intervention, a certain degree of external environmental influence—such as soil disturbance and land management activities—was unavoidable and may have caused minor data interference. Throughout the monitoring period, natural vegetation (e.g., herbaceous weeds) was manually removed during each growing season to prevent confounding effects from secondary plant species. Thus, Salix matsudana × alba remained the sole vegetation cover in all plots. No intensive human management was applied during the tree growth to minimize such effects. Instead, soil improvement was primarily driven by the plant’s intrinsic physiological regulation mechanisms.

2.3. Soil Sampling and Physicochemical Property Analysis

From October 2019 to October 2024, soil sampling was conducted annually during the same season (October). All sampling was performed between 11:30 and 14:30 each day to minimize the impact of diurnal variation on soil physicochemical properties. Using a grid layout method, six sampling points were randomly selected within each salinity treatment area. Sampling points were randomly distributed within each treatment plot while maintaining a minimum distance of 1 m from the base of tree trunks to avoid rhizosphere and canopy shading effects. This ensured that soil samples reflected general plot conditions rather than localized root-zone effects. At each point, a 100 cm³ stainless steel cutting ring was used to collect soil samples from the 0–20 cm soil layer. This sampling depth refers to the research of Poeplau et al., which used 0–20 cm or 0–30 cm as the main sampling and analysis depth, which is suitable for studying the impact of vegetation restoration on soil [17]. During vegetation restoration, the input of litter, root growth, exudation, and the resulting microbial activity is concentrated in this layer, leading to significant changes in key physicochemical properties such as soil organic matter (SOM), nutrient content, and aggregate stability. Therefore, monitoring the 0–20 cm layer effectively captures the early and primary effects of ecological restoration measures on soil improvement.

Soil bulk density (BD) was determined using the cutting ring method to calculate the dry weight per unit volume of soil. Soil pH was measured with an FE28-Standard 3000 pH meter in a 1:2.5 (soil: water, w/v) suspension. Electrical conductivity (EC) was measured using a Bante950 conductivity meter in a 1:5 (soil: water, w/v) extract. Total salinity (TS) was determined using the residue evaporation method, with weighing performed on an ME204/02 electronic balance. Soil organic carbon (SOC) was determined using the external heating potassium dichromate oxidation method with titration. The SOC values were then converted to SOM by applying the conventional Van Bemmelen factor of 1.724 [18]. Total nitrogen (TN) was determined using a fully automatic Kjeldahl apparatus (K1100, FS305). Total phosphorus (TP) and total potassium (TK) were measured using an inductively coupled plasma optical emission spectrometer (ICP-OES, Avio 200) after digestion with an H₂SO₄-HClO₄ acid mixture. Although TP and TK represent total nutrient pools and may not fully capture short-term bioavailability changes, they were used to evaluate elemental accumulation trends over the five-year period. For heavy metals, arsenic (As) and mercury (Hg) were measured using an atomic fluorescence spectrometer (AFS-8520) after digestion. Zinc (Zn), chromium (Cr), nickel (Ni), copper (Cu), cadmium (Cd), and lead (Pb) were analyzed using an inductively coupled plasma mass spectrometer (ICP-MS, iCAP RQ) following digestion with an HNO₃-HClO₄ acid mixture. Although only total heavy metal concentrations were measured in this study, future research should incorporate assessments of mobile or bioavailable forms to better evaluate ecological risks and plant uptake potential.

2.4. Construction of Soil Quality Index

To comprehensively evaluate the effects of soil improvement under different salinity gradients and planting densities, this study constructed a soil quality index (SQI) using principal component analysis (PCA). Constructing SQI based on PCA is a widely adopted method [19], as it facilitates objective indicator selection and weight determination while effectively reducing subjectivity.

First, sixteen soil physicochemical indicators were selected, including soil organic matter (SOM), pH, electrical conductivity (EC), total salinity (TS), bulk density (BD), total nitrogen (TN), total phosphorus (TP), total potassium (TK), and eight heavy metal elements: zinc (Zn), chromium (Cr), nickel (Ni), copper (Cu), cadmium (Cd), lead (Pb), arsenic (As), and mercury (Hg). All indicators were normalized to 0 and 1 before PCA to eliminate dimensional inconsistencies. During the normalization process, the scoring direction of each indicator was determined based on its relationship with soil quality [20]. For positively correlated indicators (e.g., SOM, TN, TK, TP), min-max normalization was applied as (X − min)/max − min); for negatively correlated indicators (e.g., pH, EC, TS, BD, and heavy metals), reverse normalization was applied as (max − X)/(max − min), ensuring that higher normalized values consistently indicate better soil quality.

Subsequently, PCA was performed on 75 sets of average values derived from sixteen indicators (3 salinity gradients × 5 years × 5 planting densities), and principal components were extracted to ensure that the cumulative explained variance exceeded 80%. Indicator weights in the composite SQI were calculated based on the loading values, and the variance of each principal component was explained.

Finally, each indicator’s range-normalized value was multiplied by its corresponding weight, and the weighted sum was computed to obtain the SQI score for each treatment unit.

The SQI was calculated using the following formula:

$$\mathrm{S}\mathrm{Q}\mathrm{I}=\sum _{i=1}^n(W_i\times N_i)$$

(1)

where $W_i$ is the weight of the ith indicator, and $N_i$ is the range-normalized value of the ith indicator.

The SQI constructed using this method comprehensively reflects changes in physical, chemical, and environmental quality of soil, providing a quantitative tool for evaluating the effectiveness of different treatment strategies.

2.5. Random Forest Model

This study employed the random forest (RF) model for predictive analysis to further evaluate the predictability of SQI and the rationality of its construction method. Sixteen soil physicochemical properties were used as independent variables, and the SQI score constructed via PCA served as the dependent variable.

The dataset consisted of 450 raw samples of all soil indicators, which were randomly divided into a training set (80%) and a validation set (20%) at an 8:2 ratio. RF modeling was implemented in Python (version 3.10) using the scikit-learn library (version 1.2.2). Model parameters were set as follows: the number of trees (n_estimators) was 500, the maximum depth of the trees (max_depth) was 20, and the random seed (random_state) was fixed at 42 to ensure reproducibility.

After training, model performance was evaluated on the validation set by calculating the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Additionally, feature importance values were extracted to identify key variables that dominate SQI prediction.

2.6. Statistical Analysis

All statistical analyses were performed using R software (version 4.1.0, R Core Team, 2021) and Python (version 3.10). To examine the variation in soil physicochemical properties under different salinity gradients and planting densities, a two-way analysis of variance (ANOVA) was conducted, followed by Tukey’s honestly significant difference (HSD) test for post hoc multiple comparisons. A p-value less than 0.05 was considered statistically significant. The ANOVA results were used to determine differences among planting densities within the same year and salinity level (indicated by uppercase letters) and differences among salinity levels within the same year and treatment (indicated by lowercase letters).

Pearson correlation analysis was employed to explore the correlations among soil physicochemical indicators, and a heatmap was used to visualize the correlation matrix. This revealed the strength and direction of relationships among variables under different treatment conditions.

PCA was applied to reduce dimensionality and select key indicators in constructing SQI. PCA was performed on 75 sets of standardized mean data, and principal components PC1-PC5 were extracted to ensure that the cumulative variance explained exceeded 85%. Indicator weights were calculated based on a weighted combination of component loadings and explained variances, which were subsequently used for SQI score computation. Before PCA and SQI calculation, all soil physicochemical indicators were normalized to a 0–1 range using min-max normalization to eliminate differences in measurement units.

3. Results

3.1. Analysis of Soil Physicochemical Properties

This study systematically measured and analyzed the five-year variation in soil physicochemical properties under different salinity gradients and planting densities. The results showed that various indicators exhibited degrees of temporal change and significant differences across years, treatments, and salinity conditions. The findings include trend graphs of non-heavy metal indicators (Figure 2) and a tabulated summary of heavy metal dynamics (Table S2).

All indicators were classified by salinity level (low, medium, high) and by treatment combinations of year × planting density (from 2 × 2 m to 4 × 4 m). Error bars in the bar charts represent standard deviations. According to significance analysis, uppercase letters in the figures indicate differences among planting densities within the same salinity level and year (p < 0.05). In contrast, lowercase letters represent differences among salinity levels within the same year and treatment (p < 0.05).

Overall, soil indicators varied by salinity gradient after five years of Salix matsudana × alba cultivation. SOM content showed a consistent upward trend across all salinity levels, with more pronounced increases in the low and medium salinity zones. Although the increase in SOM in the high salinity zone was relatively smaller, it still indicated a positive trend, with the 2 × 3 m planting treatment outperforming others. Soil pH exhibited an overall decreasing trend, suggesting a shift toward neutral pH levels. This decline was more evident in the medium and high salinity zones, while changes in the low salinity zone were more stable. Among treatments, denser plantings (2 × 2 m, 2 × 3 m) were more conducive to reducing pH. TN content increased gradually over time, particularly in the low and medium salinity zones, but exhibited greater fluctuation in the high salinity area. In the fifth year, some high salinity plots observed a slight decline in nitrogen content. TP content remained relatively stable, with slight increases in the low and medium salinity zones and little change in the high salinity area. Overall, TP was less affected by planting density, with salinity gradient being the dominant factor. TK content showed limited variation over the five years, with slight increases in the low and medium salinity zones and stability in the high salinity zone. Differences among treatments were minor. EC and TS showed an overall decreasing trend. The reduction was more pronounced in the high salinity zone, and high-density planting treatments exhibited greater salinity reduction effects in the medium salinity zone. BD declined overall, with the most significant decrease occurring in the first two years after planting, followed by stabilization. The improvement in BD was greatest in the low salinity zone, with a more substantial overall decrease compared to the high salinity area. Heavy metal concentrations in the soil showed limited changes, remaining relatively stable over the five-year period. Zn and Cr displayed slight declines in the low and medium salinity zones. Pb exhibited fluctuations in the high salinity zone but showed no consistent accumulation risk. Additionally, the control plots established within each salinity zone exhibited no significant changes in soil physicochemical parameters over the five-year period. This stability in the background conditions further confirms that the observed improvements in the treatment plots can be attributed to the remediation effects of Salix matsudana × alba.

These results demonstrate that the five-year Salix matsudana × alba cultivation in coastal tidal flats significantly improved SOM content, pH, EC, and reduced BD without inducing heavy metal accumulation. On the contrary, it contributed to stabilizing certain elements, providing a secure foundation for future ecological utilization of tidal flat lands. The salinity gradient and planting density significantly influenced the effectiveness of soil improvement, with denser planting treatments (2 × 2 m, 2 × 3 m) showing overall superior performance.

3.2. Correlation Analysis of Soil Physicochemical Indicators

Pearson correlation analysis was conducted on sixteen soil variables to explore the interrelationships among soil physicochemical indicators. The results were visualized in the form of a heatmap (Figure 3). The heatmap displays the correlation coefficients (R-values) and their corresponding levels of statistical significance (p-values) for each indicator pair.

Overall, most soil indicators exhibited moderate to strong correlations, with several pairs reaching significant or highly significant levels (p < 0.05, p < 0.01, p < 0.001). Specifically, SOM showed significant negative correlations with TS and EC (r = −0.88, −0.86; p < 0.001), indicating that SOM accumulation tended to increase as soil salinity levels decreased. A very strong positive correlation was observed between EC and TS (r = 0.97, p < 0.001), aligning with the expected salinity classification indicators co-variation. Soil pH also showed a strong positive correlation with EC (r = 0.79, p < 0.001). Among nutrient elements, a highly significant positive correlation was found between TN and SOM (r = 0.94, p < 0.001). Additionally, TP and TK showed a weak correlation (r = 0.20), which is related to the relative stability and mobility of phosphorus and potassium in saline-alkaline soils. Correlations among heavy metal elements were generally weak, although some elements—such as Cu and Ni—exhibited a moderate positive correlation (r = 0.50, p < 0.01).

These results reveal the complex interactions among different soil indicators, providing both theoretical support and empirical data for the subsequent PCA and SQI construction.

3.3. Principal Component Analysis

To further identify the key indicators influencing soil improvement in coastal tidal flats, principal component analysis (PCA) was performed on all soil physicochemical variables. The PCA results showed that the first five principal components (PC1–PC5) cumulatively explained 82.5% of the total variance (Figure 4), thereby retaining most of the information contained in the original data. Accordingly, PC1–PC5 was selected as the basis for calculating the composite weights of the indicators.

The principal component loadings of each indicator across the different components are presented in

Table 1. Loading matrix of soil physicochemical indicators on principal components PC1–PC5.

	PC1	PC2	PC3	PC4	PC5
SOM	−0.356	−0.048	−0.009	0.008	0.1
pH	0.319	0.188	−0.057	−0.234	−0.1
EC	0.359	−0.054	−0.03	−0.053	0.001
TP	−0.276	−0.222	0.233	−0.151	0.129
TK	0.052	−0.467	0.066	0.175	0.051
TN	−0.349	−0.003	0.026	0.023	0.109
TS	0.361	−0.041	0.02	−0.022	0.002
Hg	0.099	0.185	0.127	0.717	0.117
As	0.139	0.191	0.237	−0.003	0.832
Ni	0.226	0.333	0.125	−0.292	−0.074
Cu	−0.147	0.481	0.201	−0.107	0.035
Cd	−0.169	0.217	0.462	−0.21	−0.2
Pb	0.079	−0.182	0.622	0.132	−0.352
Cr	0.215	−0.158	0.44	0.076	0.083
Zn	−0.054	0.41	−0.09	0.457	−0.256
BD	0.355	−0.065	−0.076	0.011	−0.016

. The results revealed that SOM, EC, and TS had high positive loadings on PC1, while BD exhibited a negative loading on PC1. In addition, heavy metals such as Cd, Cr, and Zn made notable contributions to PC2 and PC3.

By comprehensively considering the explained variance of each principal component and the corresponding loadings of the indicators, the composite weights of all indicators were calculated based on the PCA results. This approach provided a scientific foundation for the subsequent construction of SQI, effectively reducing the bias associated with single indicators and enhancing the accuracy and representativeness of the overall soil quality assessment.

3.4. Soil Quality Index

To comprehensively evaluate the improvement effects of Salix matsudana × alba on soil physicochemical properties under different salinity gradients and planting densities, this study applied PCA to integrate soil indicators and construct a soil quality index (SQI).

First, all indicator data were normalized using range normalization, standardizing variables with different units to a uniform scale between 0 and 1. Based on the cumulative variance explained by PC1-PC5 (82.5%) (Figure 4), the loading matrix of each indicator on PC1-PC5 was extracted (Table 1). Combined with each principal component’s variance contribution, each indicator’s comprehensive weight was calculated (Figure 5). Among them, pH, TP, Ni, Cd, and BD contributed the highest proportions to the composite weights. Finally, the SQI score for each sample was calculated using a weighted summation of the range-normalized values and their corresponding weights. This method effectively integrates multi-indicator information, improving comprehensive soil quality assessment’s scientific validity and representativeness.

To visually illustrate the trends in soil quality under different combinations of salinity gradients and planting densities, a bar chart of SQI changes over the five-year monitoring period was generated (Figure 6). The results showed that the 2 m × 2 m and 2 m × 3 m treatments in the low salinity zone achieved consistently higher SQI scores, indicating superior soil improvement performance. The 2 m × 2 m high-density treatment maintained the highest SQI levels throughout the five years in the medium salinity zone. The 2 m × 3 m treatment exhibited relatively better SQI improvements in the high salinity zone. Overall, SQI scores increased yearly with longer planting durations, suggesting that Salix matsudana × alba cultivation can continuously enhance the quality of coastal saline-alkaline soils.

3.5. Analysis of Random Forest Model

This study employed the random forest (RF) model for predictive analysis to further assess the reliability of the SQI constructed from soil physicochemical indicators. Sixteen original soil physicochemical indicators were used as independent variables, while the SQI scores derived from PCA (PC1–PC5) served as the dependent variable for model training.

The model demonstrated good predictive performance on the validation set, achieving an R² of 0.746, an RMSE of 0.030, and an MAE of 0.024 (Figure 7).

In addition, feature importance analysis was conducted to determine the contribution of each soil indicator to SQI prediction (Figure 8). The results showed that SOM accounted for the highest importance (31.5%), followed by TS, which contributed 17.3%.

4. Discussion

4.1. Analysis of Changes in Soil Physicochemical Properties and Their Mechanisms

This study demonstrates that after five years of Salix matsudana × alba cultivation, the physicochemical properties of coastal tidal flat soils were significantly improved, particularly in the low and medium salinity zones, and the overall soil environment was optimized, laying a foundation for the ecological restoration of tidal flat soil functions.

As shown in the results, SOM increased markedly, EC and TS decreased significantly, pH shifted toward neutrality, and BD showed a notable downward trend. SOM exhibited a continuous upward trend over the five-year planting period, with the most substantial increases observed in the low and medium salinity zones. Previous studies have indicated that willow species significantly enhance soil microbial activity and organic matter accumulation through increased root biomass and the continuous release of root exudates, constituting a primary mechanism for improving SOM during ecological restoration [21]. EC declined year by year, reflecting the gradual alleviation of soil salinity stress. Qju et al. noted that salt redistribution in the soil profile could be promoted by drainage processes and soil moisture movement, which are influenced by plant evapotranspiration. Although their study focused on irrigation conditions, similar mechanisms may occur under vegetation cover, where transpiration-driven water movement facilitates vertical salt leaching and alters EC distribution [22]. The trend in TS mirrored that of EC, showing a significant decrease, particularly in high salinity areas. Research has shown that salts can migrate from surface to subsurface layers under vegetation-driven hydrological processes—particularly when evaporation decreases and vegetation cover increases—thus facilitating salt leaching and redistribution. Additionally, fluctuations in groundwater levels and lateral recharge significantly influence surface salt redistribution and overall reduction [23]. pH declined overall over the five-year period, transitioning gradually toward neutral. The decrease in pH may be attributed to the secretion of organic acids (e.g., citric acid, malic acid) by plant roots, which release H⁺ ions that neutralize alkaline ions in the soil, promoting the acidification of saline-alkaline soils and improving the rhizosphere environment [24]. Moreover, the increase in SOM may also enhance soil buffering capacity and play a key role in stabilizing pH. BD decreased over time, indicating a transformation from a compacted to a more porous soil structure. Prior research has shown that root activity promotes the formation and stabilization of soil aggregates through physical disturbance and the accumulation of root residues. Additionally, increased SOM enhances soil cohesion and pore distribution, thereby reducing bulk density and improving aeration and water retention [25].

Besides these key indicators, TN exhibited a gradual upward trend, particularly in the low and medium salinity zones. This increase may be closely related to root activity. On the one hand, roots can secrete low molecular weight organic substances such as sugars and organic acids that stimulate nitrogen mineralization and fixation by soil microorganisms. On the other hand, carbon input in the rhizosphere can enhance microbial community activity, thereby promoting the formation and accumulation of available nitrogen [26]. TP and TK remained relatively stable, showing minimal fluctuations over the five-year period, which may be attributed to the low mobility and strong mineral binding of phosphorus and potassium in saline-alkaline soils. Although plant-available forms might provide more sensitive indicators of nutrient dynamics, TP and TK were still included in this study to reflect broader elemental accumulation trends and to ensure consistency in long-term nutrient balance evaluation. Future research should incorporate available forms to capture bioavailability dynamics better. Sundha et al. noted that in such soils, phosphorus tends to form insoluble compounds with calcium ions, such as hydroxyapatite, which limits its release and migration [27]. Although potassium is highly mobile in plants, in soil systems, its availability is constrained by water status and colloidal adsorption, leading to relatively stable accumulation in saline environments [28].

The overall changes in Zn, Cr, Ni, Cu, Cd, Pb, As, and Hg were minor, with no systemic accumulation risks observed over the five-year period. Slight decreases in elements such as Zn and Cr were noted in low and medium salinity areas, possibly due to vegetation cover promoting soil stabilization and reducing wind erosion and leaching losses. Yan et al. found that plants enhance the formation of soil aggregates through rhizosphere processes and can adsorb or immobilize metal ions in the root zone, effectively limiting their activity and mobility in soil, thereby achieving a phytostabilization effect [29]. Other studies have shown that pH changes and soil microbial activity can significantly influence heavy metals’ mobility and bioavailability [30]. Under high salinity stress, changes in soil pH and metal ion solubility may affect the diffusion behavior of metals such as As and Pb. Overall, heavy metals did not significantly threaten soil environmental safety. It is important to note that this study measured only the total concentrations of heavy metals, which, although useful for assessing overall soil contamination levels, may not fully reflect the mobility or bioavailability of these elements.

Furthermore, the study area represents a typical coastal saline-alkaline zone, where occasional localized reverse salinization may occur. However, the overall salinity level showed a downward trend. This change may be attributed to the strong salt-absorbing capacity of Salix matsudana × alba roots and transpiration-driven salt migration and to a potential dilution effect from external water sources due to nearby agricultural activity. Nevertheless, the reliability and representativeness of the experimental data were ensured through systematic plot design and statistical control.

In conclusion, Salix matsudana × alba demonstrates excellent ecological restoration potential in coastal saline-alkaline soils. It effectively improves soil quality and reduces salinity stress, thus providing theoretical support and practical reference for future greening and sustainable utilization of coastal saline lands.

4.2. Correlation Analysis Among Soil Physicochemical Indicators

Through Pearson correlation analysis, this study revealed significant synergistic variation patterns among soil physicochemical indicators after five years of Salix matsudana × alba cultivation in coastal tidal flats, providing a theoretical foundation for understanding soil response mechanisms during ecological restoration.

SOM exhibited a significant and strongly negative correlation with both EC and TS, with high correlation coefficients and extremely significant levels. This indicates that during the planting process of Salix matsudana × alba, SOM accumulation increased as EC and TS decreased, reflecting the concurrent progression of vegetation restoration and soil quality enhancement. Setia et al. reported that elevated salinity suppresses plant growth and microbial activity, reducing organic matter input, which further supports the observed trend of SOM increase accompanying salinity decline in this study [31].

A robust positive correlation was found between EC and TS (r = 0.97, p < 0.001), a pattern consistently validated across different salinity types. Research by Ismayilov et al. also confirmed that EC and TS are closely related. EC—under a 1:5 extraction ratio—is a reliable proxy for estimating TS due to its stable linear relationship [32].

pH showed a high positive correlation with EC (r = 0.79, p < 0.001), suggesting that soil acidity/alkalinity adjusted concurrently with salinity reduction. Qadir et al. reported that high salinity environments are often accompanied by elevated concentrations of alkaline ions, leading to soil alkalization [33].

A highly significant positive correlation was also found between SOM and TN (r = 0.94, p < 0.001). This may be driven by the contribution of plant litter and root residues as stable sources of organic nitrogen, along with enhanced microbial nitrogen cycling. Previous studies have shown that in various ecosystems, the formation of stable humus depends on the co-accumulation of organic carbon and nitrogen, with a consistent C:N ratio, highlighting nitrogen’s central role in SOM accumulation and overall soil quality improvement [34]. While the current study did not directly assess the C/N ratio, the strong correlation between SOM and TN implies potential co-regulation of soil organic matter quality and nitrogen status. As C/N is a sensitive indicator of organic matter decomposition, nitrogen availability, and soil stability, future research should explicitly incorporate C/N analysis to enhance understanding of nutrient cycling and long-term soil fertility dynamics.

TP and TK showed only a weak positive correlation (r = 0.20), indicating limited synchronicity in their variation across the study area. Similarly, Tian et al. observed inconsistent spatial patterns between phosphorus and potassium, reflecting the influence of different controlling factors [35]. That study noted that potassium was positively correlated with clay content and soil depth, whereas phosphorus was governed by more complex variables—explaining their weak interrelationship.

Overall correlations among heavy metals were weak, with only a moderate positive correlation between Cu and Ni. This may result from shared regulatory factors in the soil environment, such as pH-induced changes in ion activity or increased SOM-enhancing organic ligand complexation, thereby affecting their solubility and mobility. Sauvé et al. also noted that Cu and Ni exhibit similar partitioning behavior across various soil types, with their solid-liquid partition coefficients being highly sensitive to pH and SOM levels [36].

Major soil physicochemical indicators (SOM, EC, TS, pH, and TN) displayed strong interrelationships, while TP, TK, and heavy metals exhibited relatively independent trends. These patterns suggest that salt dilution and nutrient accumulation are the key processes driving soil quality improvement during ecological restoration with Salix matsudana × alba. These correlations not only reflect synergistic soil dynamics but also provide a basis for identifying key driving indicators through multivariate analysis.

4.3. Principal Component Analysis and Key Factor Identification

To further determine the dominant factors influencing soil quality, principal component analysis (PCA) was performed on all physicochemical variables derived from these interrelated dynamics. By conducting PCA on 16 soil physicochemical indicators monitored continuously over five years in coastal tidal flats, this study effectively identified the key factors driving changes in soil quality within Salix matsudana × alba plantation areas. The PCA results showed that PC1-PC5 cumulatively explained 82.5% of the total variance, covering most information in the original dataset and providing a robust data foundation for subsequent SQI construction.

The dominant indicators were concentrated in PC1, where SOM exhibited high loading values, indicating its central role in the ecological restoration of coastal saline-alkaline soils. Previous studies have shown that increases in SOM help boost crop yield and enhance yield stability, promote soil carbon sequestration, and improve soil structure, thereby supporting ecosystem function restoration [37]. EC and TS also showed high loadings in PC1, indicating that salinity dynamics play a key role in soil quality improvement. During phytoremediation, salt absorption by plant roots and surface coverage effects can promote downward leaching of surface salts while reducing evaporative accumulation, thereby effectively lowering soil EC and TS levels [38]. pH also contributed significantly to PC1, underscoring the importance of acid-base regulation in restoring coastal saline-alkaline soils. It has been reported that root-secreted organic acids significantly influence rhizosphere pH dynamics [24], particularly under high salinity-alkalinity stress, facilitating a gradual shift of soil pH toward neutrality and improving the soil microenvironment [39]. BD exhibited a negative loading on PC1, reflecting its inverse relationship with SOM and salinity-related indicators. The decrease in bulk density may result from progressive optimization of soil structure. Haruna et al.’s review indicates that plant roots and their residues can effectively reduce BD, increase macroporosity and water infiltration rate, and improve aggregate stability [40].

In addition to the indicators loaded in PC1, heavy metals such as Cd, Cr, and Zn showed higher loadings in PC2 and PC3. Although they were not the dominant contributors to overall soil quality dynamics, their behavior under specific stress conditions still warrants attention. During phytoremediation, certain heavy metal indicators may exhibit restricted mobility or stabilization due to the combined influence of pH, organic matter, root exudates, and soil colloids [41].

In summary, the PCA results revealed that SOM, EC, TS, pH, and BD acted as dominant indicators driving soil improvement in Salix matsudana × alba plantation areas, while Cd, Cr, and Zn functioned as secondary factors reflecting environmental safety trends.

PCA clarified the relative importance of each physicochemical indicator during the remediation of saline-alkaline tidal soils with Salix matsudana × alba and provided a scientific basis for SQI construction. By identifying key indicators through PCA, effective and integrative evaluation of soil quality changes under complex multi-indicator conditions can be achieved, thereby reducing information redundancy and enhancing the stability and representativeness of the evaluation system.

4.4. Dynamic Analysis of Soil Quality Index

The SQI constructed based on PCA systematically reflected the dynamic changes in comprehensive soil quality over the five-year cultivation period of Salix matsudana × alba in coastal tidal flats. Overall, SQI scores across all treatment combinations showed a continuous upward trend throughout the five-year monitoring period, indicating that vegetation restoration effectively promoted the overall improvement of soil physicochemical properties. Notably, SOM and nitrogen exhibited rapid accumulation during the early post-afforestation stage, followed by stabilization over time—a process influenced by land-use history, climatic conditions, and tree species selection [42].

Significant spatial variation in SQI dynamics was observed across different salinity gradients. The low salinity zone exhibited the highest baseline SQI levels and the greatest improvement over five years, with high-density treatments (2 m × 2 m and 2 m × 3 m) significantly outperforming other planting configurations. Lower initial salt loads in this zone facilitated root expansion and biomass accumulation, thereby accelerating salt dilution and soil quality enhancement [43]. Salix matsudana × alba displayed strong salt tolerance in the medium salinity environment, effectively alleviating salinity stress and promoting synchronized optimization of soil properties. Dense vegetation cover improved surface water retention, reduced salt reaccumulation in the topsoil, and accelerated SOM buildup and nitrogen cycling. Rath et al. noted that salinity is a key factor influencing soil microbial community structure and function, thereby affecting the rate of organic matter decomposition and nitrogen cycling [44].

SQI scores were generally the lowest in the high salinity zone and exhibited a slower growth rate. However, moderate high-density planting (2 m × 3 m) still yielded a steady improvement in soil quality. Under high salinity stress, the pace of phytoremediation is relatively slow, but optimizing planting density and species configuration can significantly enhance plant survival and aboveground biomass, thereby facilitating salt replacement and structural improvement, ultimately promoting stable soil quality enhancement [45].

Synthesizing the results from different salinity gradients and planting densities, high-density planting was most favorable for rapid soil quality improvement in low—and medium salinity areas. In contrast, moderate planting densities in high salinity zones better balanced vegetation survival and remediation effectiveness. Based on our findings, restoration guidelines can prioritize Salix matsudana × alba in coastal reclamation projects, with spacing configurations tailored to salinity gradients to maximize remediation efficiency and biomass output. These findings validate the scientific rigor and sensitivity of the PCA-weighted SQI in assessing soil improvement in coastal tidal flats and provide empirical evidence and theoretical guidance for developing vegetation restoration and land management strategies tailored to varying salinity conditions.

4.5. Validation of SQI Using the Random Forest Model

To validate the predictability and rationality of the SQI constructed through PCA, this study employed the random forest (RF) model based on 16 soil physicochemical indicators. RF has been widely applied in digital soil mapping and soil quality prediction due to its strong nonlinear fitting capability, robustness, and ability to handle high-dimensional data. RF performs well in predicting key soil attributes, maintaining dependency structures among variables, and yielding high-accuracy results [46]. The results showed that the RF model exhibited strong predictive performance on the validation set, achieving an R² of 0.746, RMSE of 0.030, and MAE of 0.024, indicating high fitting accuracy and stability.

Feature importance analysis was further used to identify the most influential indicators contributing to SQI prediction. SOM was found to be the most important variable, accounting for 31.5% of the total contribution, followed by TS at 17.3%. Ni, EC, TN, and Cr also showed high importance scores. As a key indicator of soil quality, SOM reflects structural optimization and enhanced microbial activity. It significantly boosts crop yield and production stability in agricultural systems, making it one of the most representative indicators in comprehensive soil quality assessments [37]. As an essential metric for soil salinity stress, TS plays a vital role in evaluating plant growth conditions, ionic balance in the rhizosphere, and nutrient availability, thereby serving as a critical component in soil quality assessments. Studies have shown that TS is closely related to apparent electrical conductivity (ECa), and the two can be jointly applied to monitor soil salinity dynamics and guide spatial management in precision agriculture [47].

Heavy metals such as Ni and Cr also showed moderate importance in the RF model, indicating that while not primary drivers in coastal tidal flat remediation, they remain relevant for assessing environmental safety. This observation aligns with previous findings showing complex interactions between various soil pollutants and nutrient factors, which affect both plant uptake and ecological risk control [48]. Furthermore, the relatively high importance of EC and TN suggests that the interaction between salinity status and nutrient supply capacity plays a significant role in shaping soil quality trajectories.

In summary, the RF model confirmed the robustness and predictive power of the SQI construction approach and reinforced the PCA results by highlighting the dominant roles of SOM and TS in soil quality improvement within coastal saline-alkaline environments. These findings demonstrate that combining PCA and RF provides an effective modeling and forecasting strategy for complex soil systems, offering a practical technical pathway for evaluating ecological restoration outcomes in tidal flat salinized lands. However, as the RF model was trained on site-specific data, its predictive accuracy may not be directly transferable to other coastal zones with different soil textures, salinity profiles, or climate regimes. Future model calibration with multisite data is essential to enhance generalizability.

4.6. Research Limitations and Future Prospects

Although this study systematically evaluated the effects of Salix matsudana × alba on improving soil physicochemical properties under different salinity gradients and planting densities in coastal tidal flats and successfully constructed the SQI and its predictive model, several limitations remain that should not be overlooked.

First, the study primarily focused on changes in soil physicochemical properties without incorporating deeper ecological indicators such as microbial community structure and enzyme activity. As key regulators in ecological restoration, soil microorganisms may play roles as significant as physicochemical factors in the soil improvement process and thus warrant further attention in future research.

Second, due to the inherent uncontrollability of field conditions, some plant mortality and replanting occurred in the experimental area during the monitoring period, which may have caused local fluctuations in soil indicators. In addition, the experimental plots were adjacent to cultivated farmland, and exogenous interferences (e.g., irrigation leakage, fertilizer diffusion) may have influenced certain soil parameters. Although this study minimized such disturbances through rational plot layout and statistical controls, these factors should still be considered when interpreting the results.

Third, the SQI was constructed using a PCA-based weighting approach. Future work could explore integrating other machine learning techniques—such as support vector machines or neural networks—to optimize soil quality evaluation models further, thereby improving assessment accuracy and adaptability.

Lastly, although the RF model demonstrated the feasibility of predicting SQI from individual soil indicators, its predictive accuracy was limited by sample size and variable characteristics, and the model has yet to undergo external validation across regions or years. Future studies should expand the dataset to include samples from different regions and temporal contexts to test the model’s robustness and generalizability.

In summary, this study provides strong support for soil remediation strategies and the ecological restoration of coastal saline-alkaline systems. However, further improvements to the soil quality assessment framework are needed through broader spatial scales, longer temporal spans, and more comprehensive indicator systems to enhance both scientific validity and practical applicability.

5. Conclusions

This study systematically evaluated the soil improvement effects of Salix matsudana × alba in coastal tidal flats under varying salinity levels and planting densities over a five-year period. The results demonstrated that SOM content increased significantly while EC, TS, pH, and BD decreased, indicating a substantial enhancement in soil physicochemical conditions. These improvements were most pronounced under low and medium salinity conditions combined with high-density planting (2 m × 2 m and 2 m × 3 m), while in the high salinity zone, the 2 m × 3 m configuration showed superior remediation performance, underscoring the importance of optimizing planting configuration for effective remediation.

A soil quality index (SQI) constructed based on PCA successfully integrated multiple indicators into a comprehensive assessment framework. The SQI was further validated using a random forest (RF) model, which achieved a high prediction accuracy (R² = 0.746). Feature importance analysis identified SOM and TS as the most influential contributors to soil quality variation, highlighting their diagnostic value in future monitoring and modeling efforts.

These findings confirm the strong ecological restoration potential of Salix matsudana × alba in saline–alkaline environments and provide practical reference for the selection of adaptive planting strategies in coastal reclamation areas. Nevertheless, future studies should incorporate microbial and enzymatic indicators and extend the framework to multi-site, long-term monitoring to enhance its generalizability and ecological relevance.