Comparison of soil quality assessment methods for different vegetation eco-restoration techniques at engineering disturbed areas

Study area and experimental design

This study was conducted at the Xiangjiaba hydropower station, which is located at the convergence of the Sichuan and Yunnan provinces, in southeast China (Fig. 1). The experimental sites are positioned within the major prevention area of soil and water conservation zone along the upper reaches of the Yangtze River. The area is characterized by a subtropical monsoon climate with an annual average precipitation of 1,078 mm, with approximately 90% of the total precipitation occurring between May and October. The annual average temperature ranges from 12.0 °C to 18.1 °C. Sample sites were chosen with similar evolutionary times, but with varied common eco-restoration techniques. Four vegetative ecological restoration techniques were selected to determine which soil quality assessment method was more applicable to the disturbed slopes in the area: vegetation concrete eco-restoration slope (VC), frame beam filling soil slope (FB), thick layer base material spraying slope (TB), and external-soil spray seeding slope (SS). For comparison, the soil samples from the abandoned slag slope (AS) and the undisturbed, natural forest site (NF) were used to assess whether vegetation eco-restoration techniques had an effect on soil quality. Every sample location had planting soils from the Xiangjiaba Disturbed Area, and all of the sample sites had nearly identical climates. A brief description of the study sites is presented in Table 2.

Soil sampling and laboratory analyses

Soil samples were collected at a 0–20 cm depth with a 5 cm diameter soil sampler. Five sample plots (1 m × 1 m) were chosen from randomly selected locations to represent each vegetation eco-restoration technique, with each sample plot spaced at least 6 m apart and 1 m from the site edge (Ma et al., 2022). Visible debris, such as root materials and organic residues, was removed from each soil sample before dividing the soil sample into two sub-samples: one subsample for the microbial properties analysis that was stored at 4 °C, and one for the physical and chemical properties analysis that was air-dried and then passed through a 2 mm sieve. The following indicators were measured based on existing soil characteristic analyses of the study area (Xu et al., 2017): soil bulk density (BD), soil water content (WAT), soil porosity (POR), soil organic matter (SOM), total nitrogen (TN), available nitrogen (AN), total potassium (TK), available potassium (AK), pH, urease (URE), catalase (CAT), invertase (INV), polyphenol oxidase (PPO), microbial biomass carbon (MBC), microbial biomass nitrogen (MBN), and microbial biomass phosphorus (MBP). The total of sixteen soil indicators were measured using corresponding standard analytical methods (refer to Table 3 for details).

Soil quality assessment method

A principal component analysis (PCA) was performed to select the potential soil indicators that best represented soil functions and processes from the total data set (TDS). The minimum data set (MDS) and the revised minimum data set (RMDS) were identified by selecting factors that had an eigenvalue greater than 1 and accounted for 5% or more of the variance in the database. For each principal component (PC), indicators received a weight representing their contributions to the PC. The selected indicators included in MDS and RMDS were those with a loading value within 10% range of the maximum weighted loading of each PC. A multivariate correlation analysis was conducted when there was more than one soil indicator contained in each PC (Raiesi & Pejman, 2021). All soil indicators with no correlation found between them were retained in the MDS or RMDS. Well-correlated soil indicators were considered surplus and only the highest weighted indicator was chosen for the MDS or RMDS.

Two soil scoring functions, including linear and nonlinear membership functions, were selected to transform soil indicators in MDS into unitless scores ranging from 0 to 1. Each soil indicator’s effect on soil quality was scored in an increasing order by a “more is better” scoring curve and in decreasing order by a “less is better” scoring curve. The following equations were used to compute linear scoring curves:

(1) $${\mathrm{S}}_L=X/X_{max}$$

(2) $${\mathrm{S}}_L=X/X_{min}$$where $S_L$ is the linear score ranging from 0 to 1; $X$ is the value of the soil indicator; and ${\mathrm{x}}_{max}$ and ${\mathrm{x}}_{min}$ are the maximum and minimum of each soil indicator value, respectively, corresponding to the soils of the six sampling sites (Raiesi, 2017). The “more is better” scoring was computed by Eq. (1) and the “less is better” scoring was computed by Eq. (2). The following equation was used for nonlinear scoring:

(3) $$S_{NL}=A/\left(1+{\left(x/\overline{x}\right)}^B\right)$$where $S_{NL}$ is the nonlinear score of the soil indicator ranging from 0 to 1; $x$ is the soil indicator value; $\overline{x}$ is the mean value of each soil indicator observed among the six sampling sites; A is defined as the maximum score of $S_L$, which is 1 in this study; and B is defined as the slope of the equation, which is set as +2.5 for the “less is better” curves and −2.5 for the “more is better” curves (Raiesi, 2017; Yu et al., 2018; Wang et al., 2023b).

The values of the selected indicators were transformed into a comparative SQI through the following weighted additive (Eq. (4)) equation:

(4) $$\mathrm{S}\mathrm{Q}\mathrm{I}=\sum _{i=1}^n{\mathrm{w}}_i\times {\mathrm{S}}_i$$where SQI is the weighted additive of the soil quality indexes, ranging from 0 to 1; $n$ is the soil indicator quantity included in the data set; $w_i$ is the weight of the soil indicator; and $S_i$ is the score value (Qian et al., 2023).

Six evaluation methods were determined by various scoring functions: linear scoring based on the total data set (SQI-LT), nonlinear scoring based on the total data set (SQI-NLT), linear scoring based on the minimum data set (SQI-LM), nonlinear scoring based on the minimum data set (SQI-NLM), linear scoring based on the revised minimum data set (SQI-LRM), and nonlinear scoring based on the revised minimum data set (SQI-NLRM). The SQI reflects the effects of vegetation eco-restoration techniques on soil function and soil process (Raiesi, 2017; Yu et al., 2018), with a higher SQI value indicating better soil function. Soil quality grades were compared and a correlation analysis was conducted to estimate the effect of different ecological restoration techniques.

Statistical analysis

All statistical analyses were conducted with Statistical Product and Service Solutions (SPSS 20.0) software. The effects of soil indictors and SQIs were analyzed using one-way analysis of variance (ANOVA) and means were compared with Fisher’s least significant difference test (LSD). The relativity of soil indicators and SQIs was analyzed using Pearson correlation coefficients. Principal component analysis (PCA) was used to determine indicators included in MDS or RMDS and calculate the weights of soil indicators based on norm values.

Changes in soil quality indicators

A total of sixteen soil indicators were measured at the Xiangjiaba hydropower station as potential indicators of soil quality under different vegetation eco-restoration techniques (Table 4). Different vegetation eco-restoration techniques had different effects on soil indicators. Soil originating from the VC site had significantly higher levels of AK, URE, and MBP, and the lowest level of MBN (P < 0.05) compared to the soil from other sites. The levels of soil indicators, including TN, MBC, and MBP, at the FB site were significantly lower than those at other sites (P < 0.05). Values of SOM, TN, CAT, and INV were highest at the TB site. Soil originating from the AS site had the lowest values of WAT, POR, SOM, AN, URE, INV, and PPO, but significantly higher levels of BD, TK, pH, and MBN than the soil from other sites (P < 0.05). Soil from the NF site had significantly higher levels of WAT, POR, AN, and MBC, and significantly lower levels of BD, TK, pH, and CAT (P < 0.05) than the soil from other sites. The coefficient of variation (CV) had a minimum value of 12.21% (pH) and a maximum value of 84.35% (URE).

Soil quality index

Table 5 shows the principal component analysis results of all indicators, the soil indicators differed significantly by vegetation eco-restoration technique across the study sites. The four principal components with eigenvalues >1 explained more than 91% of the variability in the original data (Table 5). These four indicators also explained more than 90% of the variance in SOM, TN, AN, TK, pH, URE, CAT, INV, PPO, and MBC, 70% of the variance in BD, WAT, POR, AK, and MBN, and 66.5% of the variance in MBP. The variability of most soil characteristics were well described by these four components. AN had the highest loading value (0.951), followed by SOM (0.950), WAT (0.920), BD (−0.918), and POR (0.916). Loading values of AN were within 10% of the maximum weighted loading of PC1, which accounted for 47.485% of the total variability (Fig. 2). Due to the significant correlation between AN, SOM, WAT, BD, and POR (P < 0.01; Fig. 3), AN was the only variable included in the MDS and reserved for the SQI estimation. PC2 explained 22.999% of the total variance and was highly loaded with PPO (0.942) and CAT (0.892; Table 5 and Fig. 2). PPO was significantly correlated with CAT (P < 0.01; Fig. 3), but had the highest loading value of the two indicators, so only PPO was retained as the indictor of PC2. PC3 explained 13.882% of the total variance and was highly loaded with URE (0.753) and MBP (0.684; Table 5). No significant differences were found for URE and MBP, so both were selected as indicators of the minimum data set. PC4 explained 7.432% of the total variance, and INV was the only highly loaded indicator (0.661; Table 5). The final indicators retained in the MDS were AN, PPO, URE, MBP, and INV.

Similar to the TDS, the principal component analysis results of biological properties showed that three PCs were identified with eigenvalues >1 and explained more than 86% of the variability of the soil biological data (Table 5). The two highly loaded indicators under PC1, CAT and PPO, were significantly correlated with each other (Fig. 3), so CAT was selected to represent PC1 because it had the highest loading value of the two (0.946). Only MBP (0.964) and URE (0.876) were highly loaded indicators in PC2 and PC3, respectively. For chemical properties, two PCs were identified with eigenvalues >1 and explained more than 89% of the variability of the soil chemical data (Table 5). The two highly loaded indicators under PC1, AN and SOM, were significantly correlated with each other (Fig. 3), so AN was selected to represent PC1 because it had the highest loading value of the two (0.991). TK (0.672) and AK (0.671) were the two highly loaded indicators under PC2, and were both selected as indicators for the revised minimum data set. The principal component analysis results of soil physical property data showed that only one PC was retained for physical properties. All the measured soil indicators in physical properties were significantly correlated with each other (Fig. 3). Therefore, only BD (−0.993) was included in the revised minimum data set. As a result, CAT, MBP, URE, AN, TK, AK, and BD were chosen to establish the revised minimum data set.

The TDS, MDS, and RMDS soil indicators were transformed using the structurally based, linear and non-linear membership functions (Eqs. (1–3)). The type of scoring curve used was determined by the contribution of the indicator to soil function. The weight of each soil indicator was determined using the ratio of its communality to the sum of the communalities of all soil indicators in the TDS or MDS. To establish the RMDS, the soil physical, chemical, and biological properties were first given an equal weight value (0.333) to emphasize the equal importance of these three types of soil properties to the soil ecosystem (Nakajima, Lal & Jiang, 2015; Yu et al., 2023). Then, a sub-weight value of each soil indicator to soil physical, chemical, and biological properties was determined by the ratio of soil indicator’s communality to the sum of the communalities of that specific type of soil property (Yu et al., 2018). The final soil quality index for TDS, MDS, and RMDS can be described as follows:

SQI-M = 0.215 S_AN + 0.208 S_ppo + 0.216 S_URE + 0.145 S_MBP + 0.216 S_INV

SQI-RM = 0.107 S_CAT + 0.112 S_MBP + 0.114 S_URE + 0.122 S_AN + 0.107 S_TK + 0.104 S_AK + 0.333 S_BD

SQI-T = 0.061 S_BD + 0.060 S_WAT + 0.061 S_POR + 0.067 S_SOM + 0.066 S_TN + 0.067 S_AN + 0.065 S_TK + 0.061 S_AK + 0.062 S_pH + 0.068 S_URE + 0.067 S_CAT + 0.068 S_INV + 0.065 S_PPO + 0.066 S_MBC + 0.052 S_MBN + 0.045 S_MBP

Changes in soil quality index

The SQI values under different vegetation eco-restoration measures varied from 0.37 to 0.77 for SQI-LT and from 0.23 to 0.63 for SQI-NLT in the site order of TB > VC > NF > FB > SS > AS (Fig. 4). The SQI value of the AS site was significantly lower than that of the other sites. The SQI values of the artificial vegetation eco-restoration sites were 48% to 106% higher for SQI-LT and 78% to 174% higher for SQI-NLT than the SQI values of the AS site. These results indicated that the artificial vegetation eco-restoration measures applied in the Xiangjiaba hydropower station have markedly improved soil quality. The range of the SQI values across the sites was 0.16 to 0.76 for SQI-LM, 0.07 to 0.70 for SQI-NLM, 0.38 to 0.81 for SQI-LRM, and 0.25 to 0.64 for SQI-NLRM. The positive relationships (Fig. 5) among the six SQIs indicate that all six SQIs can both sensitively and accurately represent and quantify the effects of different vegetation eco-restoration measures on soil quality. The regression analysis between SQI values based on the data sets of different soil indicators (Fig. 6) showed that both linear and nonlinear scoring functions yielded similar SQI values based on MDS and RMDS compared to SQI values based on TDS, and a high correlation was observed between TDS and RMDS for both the nonlinear and linear scoring functions (Figs. 5, 6).

Effect of vegetation eco-restoration techniques on soil quality indicators

The responses of soil quality indicators to different vegetation eco-restoration techniques were not consistent (Table 4), and contradictory results suggest that different restoration measures have complex influences on soil indicators (Celik et al., 2021). The highest WAT value was found in the soil from the NF site, while the lowest WAT value was found in the soil from the AS site (Table 4). This is probably due to difference in species composition at different sites. The dominant species of the NF and FB sites were trees, whose canopy can effectively shield the soil from solar radiation and reduce water evaporation. The vegetation community species richness and vegetation density at the VC and SS sites were higher than those the AS site, leading to good surface water retention at the VC and SS sites, but low water retention and storage capacity at the AS site. In contrast to the AS site, the artificial vegetation eco-restoration sites had significantly increased SOM values (Table 4). According to the classification criterion of soil nutrients, the SOM levels at the artificial vegetation eco-restoration sites were at medium levels or slightly above, but the AS site was considered SOM-deficient (Xu et al., 2017). Artificial disturbance has a negative effect on SOM, while artificial vegetation eco-restoration techniques have a positive effect on SOM. Targeted artificial control measures and improved soil management levels can effectively promote the process of vegetation restoration in disturbed areas (Xu et al., 2017).

The URE values of the VC and TB sites were significantly higher than those at other sites (Table 4). This is mainly because the dominant species of these two sites belong to Leguminosae, and thus the soil of these two sites has strong nitrogen availability. Many studies show a positive relationship between soil nitrogen supply capacity and urease activity (Jin et al., 2009; Jahangir et al., 2021; Zeng, Perry & Toyota, 2023). The lower AN value at the AS site was also considered to be an important factor affecting the soil URE activity of the AS site. The significantly lower CAT value at the NF site compared to other sites may be due to the restrictive effect of excessive soil water content on microbial growth, which exceeded the activation of CAT activities (Hallaj et al., 2022; Gui et al., 2023). Soil invertase closely participates in the circulation of organic carbon, and soil invertase activity can reflect the soil’s capacity to transform organic carbon (Hu et al., 2023). Soil polyphenol oxidase can degrade phenols produced by plant lignin and detoxify soil (Toscano, Colarieti & Greco, 2003). The INV and PPO values were significantly higher at the VC, FB, TB, and SS sites than at the AS and NF sites. These results indicate that the artificial vegetation eco-restoration measures discussed in this study improved soil remediation and the soil’s capacity to transform organic carbon. The soil biological metabolic activity at the AS site was not exuberant, with low humus, which may have resulted in low organic carbon transformation efficiency and a low INV value. Research has shown that polyphenol oxidase activity is relevant to soil oxygenation and moisture status (Ankegowda et al., 2022). Compared to the AS site, the artificial vegetation eco-restoration sites and the NF site in this study were both considered oxygen and water sufficient. The reason INV and PPO values of the NF site were significantly lower than the artificial vegetation eco-restoration sites may be because the acidic soil environment was unfavorable for the catalyzed reaction of these two enzymes (Lin et al., 2022).

The MBN values at the FB and VC sites were lower than those at other sites and showed no significant difference. This result suggests that the FB and VC sites had low levels of nitrogen transformation and utilization, making soil microbial reproduction slower at these two sites. Though MBP has a longer transformation than MBN, MBP is an extremely active component in the soil organic phosphorus fraction (Sun et al., 2013). The MBP results showed that phosphorus transformation level was one of the main factors affecting SQ at the FB site. On the whole, soil microbial biomass carbon content at the four artificial vegetation eco-restoration sites was lower than at the NF site. Although the soil organic matter at artificial vegetation eco-restoration sites was found to be at a medium level or slightly above, the activity of soil organic carbon at these sites need to be improved (Khadem et al., 2021).

Comparison of soil quality indexing methods

The total data set containing soil physical, chemical, and biological indicators can more comprehensively present the results of the soil quality assessment, but with a heavy computational burden, and the correlations among these indicators may cause information overlap (Fig. 3). The minimum data set, which reduces the workload of the soil indicator measurements without losing soil quality information, is widely used for soil indicator selection when assessing the soil quality of different land uses and restoration treatments (Raiesi & Pejman, 2021; Zhou et al., 2020; Lu et al., 2023). However, this approach may exclude the limiting soil indicators of some sampling sites when selecting soil indicators from the total data set by PCA (Liu et al., 2018). Some important soil indicators representing the soil physical, chemical, or biological properties may be omitted just because they were not highly weighted in any of the selected individual PCs (Yu et al., 2018). Therefore, a revised minimum data set approach was adopted in this article to ensure the soil indicators contained in the minimum data set represented the soil biological, chemical, and physical properties. Three biological soil indicators (CAT, MBP, and URE), three chemical soil indicators (AN, TK, and AK) and one physical soil indicator (BD) were retained in the revised minimum data set, whereas only four biological soil indicators (PPO, URE, MBP, and INV) and one chemical soil indicator (AN) were retained in the minimum data set (Table 5). The revised minimum data set equally and comprehensively considered all soil properties, thereby improving the discrimination of the weighted additive soil indicator and highlighting the importance of all soil properties (Yu et al., 2018).

According to the correlation and regression analysis (Figs. 5, 6), the accuracy of SQI values based on the revised minimum data set was superior to the accuracy of SQI values based on the minimum data set. The Nash efficiency coefficient ( $E_f$) and the relative deviation coefficient ( $E_R$) were then calculated to compare the accuracy of SQI values based on MDS and RMDS (Lin et al., 2023). The $E_f$ value of SQI-NLRM was higher and the $E_R$ value of SQI-NLRM was lower than that of the other soil indicator data set approaches (Table 6). These results indicate that nonlinear scoring based on the revised minimum data set (SQI-NLRM) can distinguish the influence of vegetation eco-restoration measures at the Xiangjiaba hydropower station on soil quality more sensitively and accurately than other methods.

Evaluation of soil quality under different vegetation eco-restoration techniques

Vegetation eco-restoration techniques significantly affected soil quality in the disturbance areas at the Xiangjiaba hydropower station. SQI values differed significantly between the artificial vegetation eco-restoration sites and the unrestored disturbance site (Fig. 4). Using the soil quality graded standard established by many scholars on different land use types, such as mining wastelands and soil tillage management, the soil quality of disturbed areas at the Xiangjiaba hydropower station could be divided into five groups, as shown in Fig. 7 (Zhao et al., 2020). Based on the SQI-NLRM, high SQI values were observed at the TB, VC, and NF sites, with a medium SQI value observed at the FB. The artificial restoration base materials of the TB and VC sites have strong anti-erosion abilities. The dominant species of these two sites belong to Leguminosae (Zhao et al., 2021). The plant root of Leguminosae can secrete large amounts of sugars and amino acids, resulting in a positive impact on soil microorganism and nutrients (Yao et al., 2022). As a result of the effect of these factors, the SQI values of the TB and VC sites were greater than the SQI values of other sites. The trend in the range of values for SQI-NLRM was similar to that of SQI-NLT, indicating that SQI-NLRM is suitable for accurately assessing the soil quality of vegetation eco-restoration sites.

The applicability of the revised data set for soil quality assessment under different vegetation eco-restoration techniques at engineering disturbed areas was proven in this study. The SQI calculated based on RMDS reflects the combined response of seven important indicators to different vegetation eco-restoration measures: CAT, MBP, URE, AN, TK, AK, and BD. The contribution ratios of these seven soil indicators contained in the revised minimum data set to the soil quality index were then calculated to analyze the individual indicator response to different vegetation eco-restoration measures (Fig. 8). The contribution of the seven soil indicators significantly varied among different vegetation eco-restoration measures (Fig. 8). The contribution ratio of BD to the SQI of different sites ranged from 30.20% to 46.04%, with an average contribution ratio of 37%, which was the largest contribution indicator to SQI in the study area (Fig. 8). Wang et al. (2023b) found that SOM contributed the most to SQI when evaluating soil quality in grassland and Li et al. (2019) found that alkali-hydrolyzable nitrogen contributed the most to SQI in highly productive farmland during the wheat phase of a wheat-maize cropping system. However, in this study, BD was identified as the largest contributor to SQI. This difference may be attributed to variations in plant types, as the previous studies focused on restoring grass and wheat, and the vegetation community of the study area in the present study was more complex after a long period of ecological restoration, with a combination of trees, shrubs, and grasses. This vegetation community led to an increase in rooting and microbial activity, resulting in significant changes in BD and an improvement in soil structure (Nyirenda, 2020; Gu et al., 2019). Soil bulk density is an important and sensitive soil quality indicator and is closely connected with soil functions, such as the soil nutrient cycle, soil nutrient storage, and soil aggregate degree (Xie et al., 2017). Soil bulk density also has a significant effect on plant root growth and physiological activity (Zhang et al., 2023a). Relatively high bulk density is considered to be the reason why the SQI of the AS site was much lower than the SQI of other sites. The susceptibility of soil bulk density to different vegetation eco-restoration measures and its high contribution ratio to soil quality in disturbed areas of Xiangjiaba hydropower station indicate that soil bulk density is an indicator that can potentially reflect effect of vegetation restoration measures on soil quality. More attention should be paid to soil bulk density in vegetation eco-restoration work.

The main objective of soil quality assessments is to accurately investigate the condition of the soil, find out whether problems exist, and then provide the basis for soil quality improvement. In engineering applications, the total data set can comprehensively express soil quality information of the study area and has referential value for accurate understanding of the soil quality of study area, but the total data set usually contains multiple indicators that have no ability to reflect or control the effects of vegetation restoration techniques. The soil indicators retained in the RMDS soil quality evaluation method can comprehensively express soil quality information of the study area, and the evaluation results can reflect the impacts of different vegetation eco-restoration measures on soil quality. The RMDS is recommended for assessing soil quality during vegetation eco-restoration in disturbed areas, such as the Xiangjiaba hydropower station and other areas with similar habitat characteristics. Simplifying the total data set can effectively and accurately guide the vegetation restoration effect monitoring and control work of disturbed areas. However, the selected soil quality evaluation method should depend on the actual requirements and purpose of the evaluation in engineering practice. Additionally, given the potential influence of seasonal variations and meteorological conditions on plant growth and the outcomes of ecological restoration, temperature and precipitation should be further examined as variables in future studies.