The research area is located in the slope within the transmission line corridor in Wanzhou District, Chongqing City, China (30°48’N, 108°24’E), which is in the Three Gorges Reservoir area. The region is characterized by mountainous terrain with elevations ranging from 150 to 1,200 m above sea level. The area has a subtropical monsoon climate with a mean annual temperature of 18.3°C and annual precipitation of 1,200 mm, concentrated from May to September. Intense rainfall events in this region affect both slope stability and ecological restoration outcomes (Li et al., 2022).

The experimental site (S1) is located at 580 m elevation within a 500 kV transmission line corridor. As shown in Figure 1, the slope has a height of 12 m, length of 18 m, and a 45° inclination facing southeast. The soil profile comprises three layers: topsoil (0–0.3 m), weathered sandstone (0.3–2.5 m), and bedrock (>2.5 m). Prior to treatment application, the slope exhibited sparse vegetation and visible erosion, typical of disturbed areas in transmission line corridors. Site conditions are representative of slopes requiring ecological restoration in the Three Gorges Reservoir region, enabling meaningful comparison of restoration methods (Li et al., 2022).

The study adopted a Randomized Complete Block Design for comparing the effectiveness of four ecological restoration methods on the slope supporting the transmission line. The experimental treatments included: (1) vegetation concrete (VC), a porous cement-based substrate designed to provide mechanical stabilization while facilitating vegetation growth (Kim and Park, 2016); (2) hydroseeding (HS), where seeds, mulch, fertilizer, and water were applied for rapid surface coverage; (3) ecological bags (EB), using biodegradable geosynthetics as the growth substrate (Guo et al., 2025); and (4) natural recovery (NR), the control treatment with no human intervention. A total of 12 experimental plots were established on the slope.

Treatment details are listed in Table 1. The vegetation-concrete mixture was made up of Portland cement, locally obtained top soil, organic compost, and water retention polymers, which were combined in relation to optimized ratios for porosity and mechanical strengths (Kim and Park, 2016). Hydroseeding involved a mixture consisting of indigenous grasses and legume seeds, with wood fiber mulch and tackifier, applied at optimized pressure. Ecological bags were made from degradable jute cloth, with soil/compost mix, vegetated with indigenous shrubs two months before their final installation in the field (Guo et al., 2025). All vegetation restoration treatments were done in May 2023 using optimized slope preparations, with surface cleaning and minor grading for similarity in starting points.

The experimental design, presented in Figure 2, reflects the random spatial distribution of treatment plots (10 m × 10 m each), which reduces the impact of topographic-Edaphic gradients on the slope area. Random number-based allocation for treatment plot identification avoids any systematic error in the experimental design. Such a well-balanced experimental design helps in making appropriate comparisons between restoration levels using statistical analyses, accounting for heterogeneity in natural settings with suitable replication levels (Eab et al., 2015). Altogether, the design enables the evaluation of geotechnical stability measures, in addition to ecological metrics, in the same settings, making it easier to mitigate any factors considered in making comparisons in other restoration research studies.

A comprehensive monitoring system was established to evaluate geotechnical stability and ecological function throughout the 24-month study period. The monitoring system, listed in Table 2, comprises two different main sets: slope stability indicators and ecological function parameters. Slope stability was assessed using four main parameters. Surface displacement was measured monthly using a total station system to detect slope movement and provide early warning of potential instability (Bordoloi and Ng, 2020). Soil shear strength was measured using direct shear tests on intact samples extracted at 6-month intervals from the 0–30 cm depth, a key parameter for understanding root-soil mechanical integration (Jian et al., 2024). Surface erosion rates were measured using erosion pins and sediment traps at representative sites, with measurements conducted after each rainfall event to capture erosion dynamics. Root pull-out strength was measured for roots with diameters of 1–5 mm, acknowledging the heterogeneity in mechanical properties among species (Xian et al., 2025) and potential seasonal variations related to phenological cycles (Ji et al., 2025).

Ecological performance was assessed using integrated vegetation and soil measurements. Vegetation cover was measured monthly using UAV photography and image analysis, enabling rapid assessment across large areas (Li et al., 2023). The integrated UAV system helps in making an accurate assessment of vegetation establishment dynamics (Xu et al., 2022) and allows the extraction and calculation of vegetation cover in complex terrain (Chen R. et al., 2023). Species diversity was measured quarterly using quadrat sampling (1 m × 1 m), with diversity metrics calculated using the Shannon and Simpson indices. Aboveground biomass was measured semiannually using destructive sampling followed by oven-drying. Soil parameters, such as organic matter, available nutrients (nitrogen, phosphorous, potassium), and soil pH, measured according to conventional laboratory protocols on a 3- to 6-month cycle, enable a comprehensive evaluation of system performance with regard to multiple engineering stability and ecological function criteria (Yazdani et al., 2024).

Data collection activities were done over a period of 24 months from May 2023 to April 2025, covering two entire life cycles for the observation of seasonal variations in terms of stability and ecological factors. The period for monitoring activities is quite long since it is meant to test the long-term efficiency of the restoration treatment considering that soil bioengineering systems need time for the development of root systems (Bischetti et al., 2014). The pattern for carrying out monitoring activities is dependent on the nature of parameters being measured, with monthly monitoring for dynamic factors like surface movement, vegetation cover, among others, and semi-annual monitoring for parameters like shear strength, root tensile strength, and biomass accumulation.

Standardized protocols were followed in each measurement to allow for comparison between treatments. Displacement measurements on the surfaces were carried out with a Leica TS16 total station with an accuracy of 2 mm, with control points being fixed on stable rocks away from the treated area. Soil shear strength measurements were done with a portable strain control direct shear tester (ZJ), with intact specimens being made up of soil and roots (10 cm × 10 cm × 5 cm in dimension). The root tensile strengths were measured using a digital force gauge with a resolution of 0.01 N, with mechanical clamps following a pull-out test procedure. The vegetation maps were created using a DJI Phantom 4 RTK drone at an altitude of 30 m with standardized image settings regarding pixel size and light exposure. The simultaneous observation of root systems together with soil physical properties is an essential aspect for the understanding of slope stability mechanisms (Osman and Barakbah, 2006; Ghestem et al., 2011).

Additionally, quality control measures were adopted that involved the following: (1) calibration before each sampling, (2) replicate sampling (n ≥ 3) for each destructive test, (3) sampling locations with permanent identification for replication, (4) concurrent data entry with checking, and (5) independent surveyors’ double-checking for vegetation measurements. All analyses in the four labs followed national standards with certified reference standards for quality control.

All statistics were performed using the R software (Version 4.3.1), supplemented with SPSS 26.0 for some statistics. When done, the scores were then submitted to careful preprocessing for outlier tests using the inter-quartile range test, tests for normality using the Shapiro-Wilk test, and homogeneity of variance tests using Levene’s test. Missing values, present in less than 3% of the total number of values, were replaced using multiple imputation approaches for integrity reasons.

The treatment differences on individual parameters were tested using one-way ANOVA with treatment as fixed factors and block as a random factor for incorporating spatial heterogeneity. In cases where the assumptions were violated in ANOVA, Kruskal Wallis tests were performed for comparison. Pairwise differences were tested using Tukey’s Honestly Significant Difference test with α = 0.05 significance level for contrasting differences among treatments. Changes over time were tested using two-way repeated measures ANOVA for treatment × time interaction.

Principal component analysis (PCA) was employed for dimensionality reduction and identification of dominant trends among monitoring indicators. PCA helps in the identification of uncorrelated principal components, which explain the maximum variation in the dataset, thus making it easy to visualize treatment performance in multiple dimensional spaces (Wu et al., 2022). The combination of PCA with weighting techniques helps in indexing complex systems effectively (Kurek et al., 2022).

A comprehensive assessment framework was developed by integrating the entropy weight method with the Analytic Hierarchy Process (AHP) for indicator weighting (Kurek et al., 2022; Pliego-Martínez et al., 2024). The entropy weight method quantifies information content based on data variability across treatments (Wu et al., 2022). For indicator j, the entropy value Ej is calculated as Equation 1:

where n represents the number of treatments, and pij denotes the normalized proportion of indicator j for treatment i. The entropy-based weight Wj(E) is derived as Equation 2:

where m represents the total number of indicators. The AHP method incorporated expert judgment through pairwise comparison matrices, with consistency ratios (CR) maintained below 0.10 to ensure logical coherence (Kurek et al., 2022). The combined weight for each indicator was computed using the geometric mean (Equation 3) (Pliego-Martínez et al., 2024):

followed by normalization to ensure ∑j=1mWj=1. This integration strategy balances objective data-driven assessment with domain expertise, reducing potential bias from single-method approaches. Preliminary comparison revealed weight differences below 15% between entropy and AHP methods across all indicators, validating the compatibility of both weighting schemes. The comprehensive performance score Si for each treatment was calculated as the weighted sum of normalized indicator values. Statistical significance was set at α = 0.05 for all analyses.

Geotechnical performance varied significantly among restoration treatments throughout the 24-month monitoring period (Figure 3). Surface displacement at 24 months was lowest in the VC treatment (3.1 ± 0.5 mm), representing an 83% reduction relative to the NR control (18.7 ± 2.1 mm; Table 3). The EB treatment exhibited intermediate displacement (4.5 ± 0.6 mm), while HS showed moderate stabilization capacity (7.2 ± 1.0 mm). Seasonal fluctuations in displacement were observed across all treatments, likely attributable to variations in soil moisture content, though restored plots consistently maintained lower displacement values than the control throughout the observation period.

Soil shear strength increased progressively across all restoration treatments, with substantial gains observed at 12 months reflecting root system maturation. As illustrated in Figure 3b, with more specific information in Table 3, the treatment with VC resulted in higher shear strengths (45.8 ± 4.1 kPa) at 24 months, indicating an improvement of 79% over NR (32.6 ± 3.0 kPa). EB treatment showed equally good results (52.4 ± 3.8 kPa), with values from treatment HS (48.7 ± 3.6 kPa) in between. Surface erosion measurements corroborated these stability findings, with total erosion at 24 months ranging from 1.2 ± 0.3 kg/m² (VC) to 5.4 ± 0.8 kg/m² (NR), confirming the effectiveness of engineered treatments in erosion prevention.

Root tensile strength values were found to fluctuate considerably between treatments, which is reflected in Figure 3d. The VC root had an average tensile strength of 12.8 ± 1.5 MPa, which is considerably higher compared to other treatments: 8.6 ± 1.2 MPa (HS), 10.3 ± 1.3 MPa (EB), and 4.2 ± 0.8 MPa (NR). There were highly significant treatment differences for each stability criterion (p< 0.001), with separate performance levels in each case distinguished by post-hoc tests (Table 3). The superior performance of VC across multiple indicators reflects the synergistic benefits of mechanical reinforcement from the cement-based matrix combined with biological stabilization from vegetation development. In contrast, the limited performance of NR demonstrates the inadequate self-recovery capacity of disturbed slopes without intervention.

Ecological performance differed significantly among restoration treatments over the 24-month period (Figure 4). Vegetation cover percentage showed varying establishment levels over time according to each restorative treatment strategy. The VC restorative treatment had better coverage consistency with 82.3 ± 3.8% at the 24th month, followed by treatment EB with 78.6 ± 3.5%, followed by treatment HS with 72.5 ± 4.2%, while treatment NR covered only 32.4 ± 4.5%, which were statistically significant (Table 4; F = 125.8, P< 0.001). There were also notable differences in vegetation cover percentage over time with marked fluctuations in the seasons, particularly in treatment HS in which winter caused drastic cover percentage decline followed by marked recoveries during the succeeding growing seasons, thereby indicating the sensitive nature of herbaceous vegetation communities in relation to different environmental factors, thus highlighting the importance of treatment in relation to the prevailing regional climatic factors.

Species diversity analyses yielded notable findings regarding engineered restoration systems. The HS treatment achieved the highest Shannon diversity index (2.15 ± 0.18), surpassing EB (1.84 ± 0.15) and VC (1.68 ± 0.14), despite having lower vegetation coverage. This may be attributed to the porous substrate structure in hydroseeding, which facilitates species colonization and enhances habitat-level diversity. Upon analyzing species diversity, treatment differences were significant (F = 28.5, P< 0.001), with NR treatment possessing the lowest species diversity (1.35 ± 0.12), even with seed restrictions in propagation because of exposure to harsh environments even on top surfaces.

Productivity assessment using aboveground biomass analyses showed significant hierarchical differences among treatments, with VC showing the highest accumulation (1045 ± 78 g/m²), followed by EB (965 ± 72 g/m²), HS (825 ± 65 g/m²), and NR (358 ± 45 g/m²). Soil organic matter accumulation followed similar patterns across treatments, with progressively declining rates consistent with ecological succession theory.

Simultaneously, the VC treatment showed significant improvement in soil quality (23.6 ± 2.1 g/kg), appreciated at 141% from its original status, while EB achieved 21.5 ± 1.8 g/kg, HS reached 19.4 ± 1.6 g/kg, and NR showed modest enhancement (11.5 ± 1.4 g/kg). The significant differences in soil organic matter among treatments (F = 52.3, P< 0.001) highlight the importance of active intervention to accelerate pedogenesis and achieve restoration objectives within operational timeframes.

Temporal patterns of ecological indicators over the 24-month period revealed distinct developmental trajectories among the four restoration treatments (Figure 5). Vegetation cover development varied by treatment; VC exhibited rapid colonization during the first six months, reaching approximately 82% coverage before entering a stabilization phase with minimal subsequent variation. The HS treatment demonstrated high levels of variation in values over the course of the seasons, fluctuating between 65-80% due to temperature and rainfall factors, eventually reaching stabilization at 72.5% values at the end of the period. The EB treatment demonstrated high levels of stability in vegetation cover development, while the NR treatment showed limited improvement, reaching only 32.4% after 24 months.

The Shannon diversity index curves were different in terms of the species accumulation pattern in each treatment group. The species enrichment pattern in the HS treatment followed an ongoing curve for the entire experimental period to reach its maximum diversity at the end (2.15), while the others followed an asymptotic curve toward their maximum diversities (1.68 for the VC treatment and 1.84 for the EB treatment) at 18 months, respectively. The NR treatment showed delayed species accumulation, reflecting the slower pace of natural colonization without intervention.

Biomass increase and the formation of soil organic matter followed the general slowing rates for each treatment, in line with traditional models for ecological successions. The VC treatment showed the greatest rate for the increase in biomass during the establishment phase (0–12 months), then slowing down for volume increase, eventually reaching 1045 g/m². Soil organic matter development offered insightful views into treatment-driven pedogenesis, with the NR method demonstrating little progress in the first six months, then continuing with only small increases (11.5 g/kg at month 24). The delayed soil organic matter development in the NR treatment underscores the importance of active restoration interventions, as natural recovery alone is insufficient to meet restoration objectives within operational timeframes on heavily degraded transmission line slopes.

The integrated assessment using the multi-criteria evaluation method created an unequivocal hierarchical ordering with six weighted criteria (Figure 6; Table 5). The criteria weight factors were derived from the proposed assessment criteria, allocating 38% for stability criteria (shear strength and root tensile strength), while ecological criteria were allocated a combined 62% weight, accounting for engineering needs for the restoration of the transmission line on the slope sides.

Performance comparison in multiple dimensions showed specific treatment responses (Figure 6a). The VC treatment performed better than other treatments in the structural engineering criteria with highly scored values (normalized scores > 98) for shear strength, root tensile strength, vegetation coverage, aboveground biomass, and soil organic matter; it also showed moderate Shannon index values (78.1 ± 0.9). In contrast, the HS treatment achieved the highest biodiversity score (100 ± 0.7) but moderate geotechnical scores (shear strength: 71.0 ± 1.5; root tensile strength: 67.2 ± 1.4). EB treatment performed moderately in multiple criteria with no special advantage/disadvantage for any criteria.

The scoring values resulted in the following rank: VC (99.8 ± 1.2) > EB (84.8 ± 1.3) > HS (72.8 ± 1.5) > NR (42.5 ± 1.6), with highly significant differences between treatments (F = 428.5, P< 0.001). The excellent performance in VC resulted from the integrated efforts in mechanical strengthening and biological stabilization. The EB treatment gave good scores because of its stable average scores in medium to high levels, indicating its viability in areas where well-balanced geotechnical-ecological performance is required. The average position in rank for the HS treatment resulted from its basic restrictions in structural reinforcement capabilities despite its high diversity levels. The much lower score in NR (42.5 ± 1.6) validated the unsuitability of passive restoration principles in meeting operational stability criteria in acceptable periods on heavily altered transmission line slopes.

Pearson’s Correlation analyses showed there were significant relationships between geotechnical stability parameters and ecological performance indicators for each treatment level in the time sequences (Figure 7, n = 48). The investigation pointed to mostly positive significant relationships between each pair of indicators, although the nature and significance of these relationships ranged considerably depending on the pair of parameters investigated.

Root tensile strength had the highest positive correlation with aboveground biomass accumulation (r = 0.963, R² = 0.927, P< 0.001), demonstrating that biomass accumulation is a good predictor for the mechanical reinforcement capability of roots (Figure 7b). This is in line with the direct positive relation between photosynthesis and resource allocation to roots, where higher aboveground biomass accumulation is directly associated with higher levels of root system development with higher tensile strengths. This relationship has practical implications for slope management, as vegetation biomass can serve as a non-destructive field indicator for estimating root mechanical reinforcement capacity. Similarly high positive correlations were also obtained between shear strengths and aboveground biomass (r = 0.951, R² = 0.904, P< 0.001), as well as between root tensile strength and vegetation cover (r = 0.907, R² = 0.822, P< 0.001), indicating the coupled processes within vegetation establishment and slope stabilization.

Organic matter content in the soil had a positive significant relationship with shear strength (r = 0.903, R² = 0.815, P = 0.001), indicating that soil pedogenesis is an equally important mechanism for improving geotechnical stability in addition to root reinforcement (Figure 7c). This highlights the importance of soil development as a mechanism linking ecological restoration to slope stability, potentially through aggregate formation, enhanced water retention, and improved soil cohesion. The relationship between vegetation cover percentage and shear strength (r = 0.890, R² = 0.792, P = 0.001) provided additional evidence concerning the pivotal mechanism provided by vegetation establishment. From an engineering perspective, this correlation enables UAV-based vegetation monitoring to serve as an indirect indicator for soil shear resistance changes, supporting cost-effective slope stability assessment.

Conversely, although Shannon diversity index is generally characterized by strong positive correlations with the other ecological performance metrics, it generally displayed moderate positive associations with shear strength (r = 0.496, R² = 0.246, P< 0.001), indicating a much weaker level of associations compared with other ecological metrics (Figure 7f). This pattern reflects a trade-off between species diversity and functional homogeneity in engineered restoration systems, where the treatments with maximum geotechnical capacity (primarily VC) supported high levels of system stability with rather homogeneous vegetation communities, thereby outperforming treatments with higher levels of spontaneous species immigration (primarily HS), which supported high levels of diversity at the expense of structural reinforcement capabilities. Collectively, these results thus show that while the majority of ecological stability performance metrics generally associated closely with stability improvement, diversity is a partially independent criterion requiring separate treatment in ecological restorative strategies targeting geotechnical and ecological performance objectives.