Influence of vegetation restoration on matrix structureand erosion resistance of iron tailings sites in eastern Hebei,China

Anning Wang · Qiuxian Huang · Xuehua Xu · Xiaogang Li · Yuling Li

Abstract The effects of vegetation restoration on matrix structure and erosion resistance of iron tailings were studied at dump sites in Malanzhuang, Qian’an, Hebei province, China. The restoration process involved soil spray sowing restoration model with Rhus typhina,soil and iron tailings admixture spray sowing restoration model with Amorpha fruticosa Linn. and six-hole brick restoration model with Pinus tabulaeformis Carrie`re.-Amorpha fruticosa Linn. mixed-forest, and direct restoration model with Hippophae rhamnoides and Sabina vulgaris. Results show that the composition and distribution of particles and aggregates were closely related to erosion resistance(P ＜0.05), indicating that matrix structure of iron tailings play an important role in erosion resistance. The improvement in matrix structure of iron tailings by the different restoration models was in the order of R. typhina soil spray sowing ＞A. fruticosa soil and iron tailings admixture spray sowing ＞mixed-forest six-hole brick ＞H. rhamnoides direct restoration ＞S. vulgaris direct restoration. The R. typhina soil spray sowing restoration model resulted in the greatest improvement in matrix structure of iron tailings, increasing the clay (10.6%) and large particle aggregates (18.7%) contents significantly(P ＜0.01). Simultaneously, particle population, grading conditions(Cu = 28.86,Cs = 1.65),and aggregate stability(6.02) were significantly improved. The A. fruticosa soil and iron tailings admixture spray sowing restoration model,which effectively improved particle distribution (Cu-= 8.51, Cs = 1.07), increased the number of large aggregates considerably (9.6%), thereby increasing aggregate stability (6.2). The six-hole brick model significantly increased the number of large aggregates (4.0%) and improved the stability of aggregates (6.2). H. rhamnoides direct restoration improved the stability of aggregates(5.1)but showed no other significant improvements. The effect of S. vulgaris direct restoration on matrix structure of iron tailings was not significant. Due to its dependence on matrix structure of iron tailings,the erosion resistance of R.typhina soil spray sowing restoration model was the greatest,while that of S.vulgaris direct restoration was the weakest.There was no significant difference in the erosion resistance of the other models. Overall, vegetation restoration supplemented by soil spray sowing restoration and engineering measures is superior to in situ direct vegetation restoration in the short-term. In-situ direct restoration has long-term ecological significance because of its advanced concept of near-natural restoration and the advantages of low cost,easy operation,and low secondary damage.

Keywords Erosion resistance · Matrix structure · Iron tailings · Vegetation restoration

Introduction

Iron tailings matrix refers tofine or coarse particles formed after iron ore mining, which contains a lot of iron and a little of soil (Ferreira Da Silva et al. 2005). There is no definitive classification for iron tailings. Generally, the tailings size is divided according to the principles of Soilclassification, simultaneously, the study of tailings matrix(such as structure, corrosion, nutrients, etc.) is also modeled on the general soil(Jordan 2009).Typically,iron mine tailings occupy a large area of land and mining in China entails an accumulative area of 28,800 km2, a figure that has increased at an annual rate of approximately 467 km2.Tailing waste particles are small, have strong mobility,weak water retention, fertilizer preservation capacity, and self-stability (Guo and Zhou 2006). As a result, fine sands are eroded by runoff and high winds, leading to river channel sedimentation,water pollution,and frequent floods(Mandy et al. 2009). During the rainy season, mud flows into rivers and the erosion modulus in areas near a tailing dam can be as high as 15,000 t km-2yearly, but the infiltration rate of tailing sands is only 8.5 mm min-1.Meanwhile, the sand content, erosion amount, and volume of runoff in a single episode of soil erosion can be as high as 869.7 kg m-3, 10,709.8 t km-2, and 26,291.3 t km-2respectively (Chen et al. 2015). Therefore, tailing dump sites are degraded lands with poor soil texture and an endangered biotope landscape.

In a broad sense, matrix structure refers to the quantity of particles and aggregates formed through natural binding,as well as their spatial arrangement and combination(Burri et al.2009).A favorable matrix structure can promote plant growth(Mulholland et al.1996)by increasing the mobility and accumulation rate of nutrients and moisture in pores.The appropriate degree of matrix compactness can also improve the growth rate of roots and adjust the time of stomatal movement (Hales et al. 2009). Simultaneously,matrix structure is a critical factor affecting water infiltration and overland runoff. Therefore, degraded land remediation predominantly involves remediation of matrix structure.

However, the formation of matrix structure is a lengthy process in which microorganisms and plant mechanisms are crucial. Of the two, plants have a more remarkable influence on structure modification. For example, Andruschkewitsch et al. (2014) suggested that plants can accelerate matrix structure ripening directly through root enwrapping and indirectly through litter decomposition.Consequently, vegetation restoration is the most effective and fundamental strategy for accelerating soil ripening,modifying structure, and reducing water and material loss in iron tailings (Wong 2003). As a result, investigating the effect of vegetation restoration on texture modification in iron tailings is an urgent research goal. Over the past few decades, domestic and foreign companies have gradually implemented erosion control measures in tailings through multiple vegetation restoration activities. For example, the annual erosion modulus at the Yuanjiacun iron ore mine in China was 8500 t km-2. After implementation of vegetation measures, it decreased to 1225 t km-2. Amponsah-Dacosta (2015) used bioengineering technology to comprehensively manage erosion at the Copper Cliff tailings dam in Sudbury, Canada; the rate of erosion decreased from 1000 to 257 t km-2a-1. Guo et al. (2015) reported on the runoff and sediment production processes of different bioengineering dumps at a typical large-scale openpit coal mine in China. The erosion modulus on the bioengineered slope was 63.5% lower than on the bare tailings. However, studies on erosion of tailings controlled by vegetation restoration mainly focus on external performance effects. There is very little research on erosion resistance measures controlling the inherent causes of tailings matrices erosion; therefore, tailings matrices modification and erosion management benefits of each restoration measure should be evaluated to determine the optimal vegetation restoration model and vegetation allocation.

This study evaluated five typical vegetation restoration models on the Malanzhuang iron mine tailings in Qian’an,Hebei Province by investigating the differences in matrix structure and erosion resistance of iron tailings under each model using mathematical theories.Our findings provide a scientific foundation for the restoration and reconstruction of soil and water conservation vegetation cover in iron tailing wastelands, as well as basic data for theoretical research and technological development of vegetation restoration of mine tailings.

Materials and methods

Study area

The study area is a region of steep terrain in Malanzhuang,Qian’an, in the eastern region of Hebei Province (39°51′-40°15′N, 118°26′-118°55′E). The area has a monsoon climate that ranges from warm temperate to semi-humid continental transition conditions, with an annual average temperature of 11.5 °C, an extreme minimum temperature of - 28 °C, and an extreme maximum temperature of 40 °C. Average annual precipitation is 712 mm. The original cover on the field sites prior to vegetation restoration consisted of tailings sand. The nutrient content was low, and the structure loose and unfavorable for vegetation growth (Li et al. 2014).

Field experiments

In 2007, Tangshan Shougang Malanzhuang Iron Ore Co.,Ltd., conducted several practical tailings vegetation restoration activities.These involved the following models:

Pinus tabulaeformis Carrie`re.-Amorpha fruticosa Linn.mixed-forest six-hole brick restoration model:first,a drain pipe was built on the slope of an iron tail mine and a cement wall was installed at the bottom. Then, six-hole brick was installed on the slope and P. tabulaeformis-A.fruticosa were planted in the holes.

Rhus typhina soil spray sowing restoration model: after the slope of the iron tailings was flattened,the external-soil with good texture was sprayed on the surface of the tailings sand; then the R. typhina forest was planted.

Amorpha fruticosa Linn.soil and iron tailings admixture spray sowing restoration model: after the slope of the iron tailings was flattened, the external-soil and iron tailings admixture was sprayed on the surface of the tailings sand;then the A. fruticosa forest was planted.

Syringa vulgaris direct restoration model and Hyppophae rhamnoides L.direct restoration model:S.vulgaris and H. rhamnoides were respectively planted directly on the sand of iron tailings.

Due to the accumulation of scattered tailings, broken terrain, and different vegetation restoration periods, it was difficult to synthesize factors such as the construction time,slope, and aspect of each model; however, all substrates were iron tailings. While ensuring a degree of consistency between the experimental plots and the surrounding environment, the five restoration models were established at August 2014; bare tailings were used as controls. Each restoration model was implemented in five plots. Geographical factors such as location, slope length, aspect and elevation were similar (Table 1).

Sample collection and analysis

Three quadrats were placed in each plot; a five-point sampling method was used to collect the upper 20-cm of undisturbed material with a borer. These samples were transported to the laboratory for analysis. The dry-wet sieving method was used to screen aggregates(Gee and Or 2002), and an MS2000 Laser Particle Sizer was used to measure particle composition. A pipette was used to measure the composition of microaggregates, and the potassium dichromate dilution method measured the organic matter content (Vance et al. 1987).

Particle structure

Particles are the basic unit of dielectric including soil,various types of industrial and mining areas abandoned such as iron tailings,copper tailings,which is an important factor that influences resistance to erosion (Rodríguez-Lado and Lado 2017). The particles can be further bonded or aggregated by physical, chemical, biochemical and biological effects toform aggregates and structure of various sizes, shapes and properties.

Particle composition

Composition is described particle content for each particle size and the fractal dimension of particles(PSD)(Liu et al.2009). The particle fractal dimension is based on the volume fractal model of Tyler and Wheatcraft (1992):

Table 1 Description of sample plots

Particle distribution characteristics

The moment method, which is used to determine particle population characteristics and particle grading conditions,is an important tool for describing particle distribution characteristics. Particle population characteristics include three indicators: particle skewness (Srk), dispersion (Se),and kurtosis(Sk),which characterize the symmetry,degree of dispersion, and concentration of particles, respectively(Xin et al. 2013). The relevant equations are:

where D10,D25,D75,and D90are the corresponding particle sizes when the cumulative percentage of particles reaches 10%, 25%, 75%, and 90%, respectively.

Particle grading conditions include the curvature coefficient (Cswhich describes the cumulative percentage of the particle shape curve) and the non-uniform coefficient

where D30and D60denote the corresponding particle sizes when the cumulative percentage of particles reaches 30%and 60%, respectively.

Aggregate structure

Aggregate is a soil structure in which a plurality of single particles are bonded together toform an aggregate(Morgan 2005; Wiesmeier et al. 2012).

Aggregate composition and stability

The aggregate stability index (ASI), calculated using the transfer matrix method(Shi et al.2010),is used to evaluate the stability of aggregates as follows:

(1) Obtain the all-level aggregate composition input matrix through dry screening:

(2) Obtain the all-level aggregate relative ratio composition output matrix Ni through wet screening:

(3) Rewrite the matrix:

(Cuwhich refers to the degree of uniform distribution of particles). The formulas for these coefficients are:

(4) Use the input and output matrices toform the matrix of particle structure retention probability X:

(5) Multiply all grain-level aggregate retention probability X values to obtain the ASI:

where Miand Niare the ith soil aggregate percentage output matrix and input matrix, respectively, which are obtained through dry screening and wet screening, and Xiis the retention rate of soil in the aggregate.

Distribution of aggregates

The standard deviation (б), peak-convex coefficient (CE),and bias factor (Cs) of the moment method is used to describe the distribution of aggregates (Inman 1952). The formulas for these are:

where u2is the second-order center moment;u3is the thirdorder center distance; and u4is the fourth-order center distance. xiis the particle size center value of the ith particle size aggregate; ¯x is the average particle size of all particle size aggregates; and fiis the corresponding percentage of the ith particle size aggregate.

Quantitative evaluation of erosion resistance

Twenty-one indices were selected to evaluate the erosion resistance of iron tailings under the different vegetation restoration models (Table 2).

Different fuzzy comprehensive evaluation methods produce different index weights, resulting in different evaluation results. In order to eliminate this discrepancy and fully understand the erosion resistance of the materials,three common methods were selected to evaluate resistance to erosion.

The method of mean square difference weights(Rae)(Wang 2017)

(1) Set different plans:

(2) Single indicator membership value:

Table 2 Erosion resistance evaluation index

(3) Mean square error decision:

Mean of random variables:

Mean square error:

Weight coefficient of a single indicator:

(4) Rae:

The coefficient of the principal component analysis (AEI)(Wang 2017)

(1) Ratio of the ith and principal component eigenvalues:

where n is the number of extracted principal components and λiis the ith principal component eigenvalue.

(2) The ith principal component score:

where aijis the weight of each component; xithe standardized value; m the number of single indicators.

(3) AEI:

The factor load of the principal component analysis(CAEI)(Dinno 2009)

(1) Single index membership value:

Ascending function optimization method:

Descending function optimization method:

(2) Principal component weight coefficients extracted by principal component analysis:

where Ciis the common factor variance and C the common factor variance sum

(3) CAEI:

Erodibility model

The formula by Shirazi et al. (1988) was used to estimate erodibility:

where Dgis the geometric average particle size

Data analysis

Data analysis used Excel 2016 and SPSS 23.0 and figures drawn using Sigmaplot 12.5.

Results and analysis

Influence of different vegetation restoration patterns on particle structure

Composition of particles

The R. typhina soil spray sowing model predominantly consisted of fine sand (43.2%) and silt (43.1%), while the contents of gravel (0.0%) and coarse sand (3.0%) were significantly less than those in other models (P ＜0.01).The silt and clay contents(10.6%) were significantly more than those in other models (P ＜0.01, Table 3). There was no significant difference in particle composition between other models and bare tailings control. The fractal dimensions of particles (PSD) also increased under each restoration model.The R.typhina soil spray sowing model showed the largest increase(2.5%),which was significantly greater than that of the controls (P ＜0.05). In summary,vegetation restoration increased the content of silt in the iron tailings to promote grain refinement. The R. typhina soil spray sowing model was a remarkable improvement while other models were less obvious.

Table 3 Composition of particles for different vegetation restoration models

Table 4 Population characteristics and grading of particles for different vegetation restoration models

Table 5 Composition of aggregates for different vegetation restoration models

Distribution characteristics of particles

Compared with bare tailings, the Se(dispersion) of all restoration models increased and Srk(soil particle skewness) and Sk(kurtosis) decreased (Table 4), indicating that the characteristics of the tailings improved after vegetation restoration. The Cuof the soil spray sowing restoration model and soil and iron tailings admixture spray sowing restoration model were greater than 5,and Cswas between 1 and 3,indicating good particle grading.Although Cu＞5 were detected in the other models, Cswas suggestive of poor grading and mechanical stability, indicating a high probability of erosion. Therefore, the characteristics of the mineral particles and the grading status were improved by soil spray sowing and soil and iron tailings admixture spray sowing models, decreasing the probability of erosion.

Influence of different vegetation restoration patterns on aggregates

Composition and stability of aggregates

Different aggregate fractions were observed under the different restoration models (Table 5). Cover material with large aggregates (＞2 mm) showed the highest restoration rates (18.7%) by the R. typhina soil spray sowing model,followed by the A.fruticose soil and iron tailings admixture spray sowing (9.6%), then the six-hole brick (4.0%), with contents significantly improved over bare tailings(P ＜0.01).Large aggregates in the iron tailings material did not change significantly from that of bare tailings(P ＞0.05)after in situ direct restoration.In summary,soil spray sowing and soil and iron tailings admixture spray sowing models worked well at increasing the number of large aggregates in the iron tailings. The direct restoration models did not significantly increase the number of large aggregates.

Studies have shown that the ASI (aggregate stability index) is an important factor in evaluating the stability of aggregates (Shi et al. 2010). The order of restoration models, from highest to lowest ASI, was the six-hole brick ＞A. fruticose soil and iron tailings admixture spray sowing ＞R. typhina soil spray sowing ＞H. rhamnoides direct restoration ＞S. vulgaris direct restoration (Fig. 1).There were no significant differences in ASI between the six-hole brick and the soil spray sowing and the soil and iron tailings admixture spray sowing models, nor between the S. vulgaris direct restoration model and bare tailings(P ＞0.05). Except for S. vulgaris direct restoration, all models significantly improved the stability aggregates.

Fig. 1 ASI values indicating aggregate stability by vegetation restoration model. RH: Rhus typhina soil spray sowing restoration model, P-A: Pinus tabulaeformis Carrie`re.-Amorpha fruticosa Linn.mixed-forest six-hole brick restoration model, AM: Amorpha fruticosa Linn. soil and iron tailings admixture spray sowing restoration model, HI: Hippophae rhamnoides direct restoration model, SA:Sabina vulgaris direct restoration model, CK: Bare tailings

Fig. 2 Distribution of aggregates for different restoration models.RH: Rhus typhina soil spray sowing restoration model, P-A: Pinus tabulaeformis Carrie`re.-Amorpha fruticosa Linn. mixed-forest sixhole brick restoration model, AM: Amorpha fruticosa Linn. soil and iron tailings admixture spray sowing restoration model, HI: Hippophae rhamnoides direct restoration model, SA: Sabina vulgarisdirect restoration model, CK: Bare tailings

Fig. 3 K values for different vegetation restoration models.RH:Rhus typhina soil spray sowing restoration model, P-A: Pinus tabulaeformis Carrie`re.-Amorpha fruticosa Linn.mixed-forest six-hole brick restoration model,AM:Amorpha fruticosa Linn.soil and iron tailings admixture spray sowing restoration model, HI: Hippophae rhamnoides direct restoration model,SA:Sabina vulgaris direct restoration model, CK: Bare tailings

Distribution of aggregates

As shown in Fig. 2, Cs(curvature coefficient) and CE(peak-convex coefficient) values of all restoration models were negative, which indicates a large degree of particle dispersion and no particular size aggregates dominant.Therefore, it is assumed that the duration of vegetation cover required to increase the distribution of aggregates is relatively long.

Fig. 4 Erosion resistance indexes (AEI, Rae and CAEI) in different vegetation restoration patterns and control (CK). RH: Rhus typhina soil spray sowing restoration model, P-A: Pinus tabulaeformis Carrie`re.-Amorpha fruticosa Linn. mixed-forest six-hole brick restoration model,AM:Amorpha fruticosa Linn.soil and iron tailings admixture spray sowing restoration model, HI: Hippophae rhamnoides direct restoration model,SA:Sabina vulgaris direct restoration model, CK: Bare tailings

Comprehensive analysis and evaluation of erosion resistance

Erodibility analysis

Erodibility refers to the sensitivity to water dispersion and suspension and can predict the intensity of future erosion(Wischmeier and Smith 1978). As shown in Fig. 3, K values(erodibility index)for all models were less than that for bare tailings. K values ranked from smallest to largest are:R. typhina soil spray sowing model ＜six-hole brick model ＜A.fruticosa soil and iron tailings admixture spray sowing model ＜H. rhamnoides direct restoration ＜S.vulgaris direct restoration. The R. typhina soil spray sowing model showed significantly smaller K than all other models (P ＜0.05). There was no significant difference between other models and the bare tailings (P ＞0.05).

Erosion resistance analysis

The AEI (erosion resistance index based on the coefficient of the principal component analysis) values for different models were significantly larger than for bare tailings(Fig. 4), which shows that restoration with vegetation improved resistance to erosion. There were significant differences in AEI among the different models (P ＜0.01).The AEI of R. typhina soil spray sowing model was significantly higher than those of other models (P ＜0.01),followed by A. fruticosa soil and iron tailings admixture spray sowing model, H. rhamnoides direct restoration, and the six-hole brick model. The S. vulgaris direct restoration model exhibited the lowest AEI.

As shown in Fig. 4, the R. typhina soil spray sowing model had the largest Rae(erosion resistance index based on the method of mean square difference weights), which was 80 times that of bare tailings.This was followed by H.rhamnoides direct restoration, A. fruticosa soil and iron tailings admixture spray sowing model, six-hole brick model, and S. vulgaris direct restoration. The restoration models were ranked from largest to smallest CAEI(erosion resistance index based on the factor load of the principal component analysis): R. typhina soil spray sowing model ＞A.fruticosa soil and iron tailings admixture spray sowing model ＞H. rhamnoides direct restoration = mixed-forest six-hole brick model ＞S. vulgaris direct restoration ＞bare tailings.

Compared to the bare tailings, site restoration with vegetation significantly improved resistance to erosion.The K values of the R. typhina soil spray sowing model were significantly lower than those of other models and its erosion resistance indexes were significantly higher(P ＜0.01).In comparison,the S.vulgaris direct restoration had the poorest performance. No significant differences were observed between the A. fruticosa soil and iron tailings admixture spray sowing model, the six-hole brick model, and the H. rhamnoides direct restoration model.Therefore, in order to clarify the differences in erosion resistance between the different models, a cluster analysis was performed. The results indicate that the particle resistance by different models could be classified into three categories: (1)the R.typhina soil spray sowing model was classed as ‘‘strong’’; (2) the S. vulgaris direct restoration model was classed as ‘‘weak’’; and, (3) all other models were assigned to the‘‘middle’’resistance category(Fig. 5).

Fig. 5 Cluster analysis of erosion resistance under different vegetation restoration models: RH: R. typhina soil spray sowing restoration model, PA: P. tabulaeformis-A. fruticosa mixed-forest six-hole brik restoration model, AM: A. fruticosa soil and iron tailings admixture spray sowing restoration model, HI: H. rhamnoides direct restoration model, SA: S. vulgaris direct restoration model

Discussion and conclusions

The results indicate that the quantity and distribution of particles and aggregates are closely correlated with erosion(Table 6),which are consistent with Peng et al.(2016).The quantity of aggregates is an important factor influencing resistance to erosion and the content of large aggregates from highest to lowest, follows the order R. typhina soil spray sowing model ＞A. fruticosa soil and iron tailings admixture spray sowing model ＞Six-hole brick model ＞H. rhamnoides direct restoration ＞S. vulgaris direct restoration. As the basic constitutional unit of aggregates,particle composition also determines its erodibility,and the AEI is negatively correlated with sand but positively correlated with clay (P ＜0.01). The R. typhina soil spray sowing model is associated with the greatest amount of clay, however the differences in clay content among the other restoration models are not statistically significant.This may be the key reason for small differences in erosion resistance among the other models (except S. vulgaris direct restoration).

Currently, research regarding the effect of matrix structure of iron tailings on erosion mostly focuses on quantity components, while little research is available on particles and aggregates distribution. Generally, particle population characteristics and grading conditions, as well as the aggregate distribution, reflect the degree of erosion.Tailings matrix with good granular composition have an extensive particle distribution range of non-uniform particles,with fine particles filling in the pore spaces formed by coarser particles, resulting in favorable mechanical properties and low probability of being flushed by rainfall.Tailings matrix has a good granular composition when it has a non-uniform coefficient of over 5 and a curvature coefficient of between 1 and 3. For surface erosion of tailings, fine particles and large grain sized aggregates are removed first by run off, which will reduce the quantity of clay and macro aggregates, resulting in a reduced coefficient of dispersion of the particles,lowered symmetry,and increased skewness and peak values. The composition of aggregates is then transformed into a negatively skewed distribution, and the peak position moves from large diameter to small diameter aggregates. Our findings reveal that the improvement of particle and aggregate distribution is a long process, and short-term restoration with vegetation will not produce substantial improvements.As a result,the effect of vegetation restoration duration on matrix structure of iron tailings should be further studied.

Table 6 Relationship between erosion resistance and particle and aggregate structure

Vegetation restoration can modify particles and aggregates in iron tailing wastelands and enhance the erosion resistance of iron tailings.This is consistent with Ma et al.(2015). The soil spray sowing is superior to the soil and iron tailings admixture spray sowing restoration models,and restoration models supplemented with engineering measures are superior to in situ direct restoration. The R.typhina soil spray sowing model achieves the best results.Firstly, the model can remarkably improve the matrix structure and microenvironment of pure tailings, resulting in a superior structure and erosion resistance of the tailing sand. Secondly, R. typhina has shallow, well-developed roots and sprouting ability. As a result, the roots can effectively immobilize the loose surface matrix structure of iron tailings, mechanically reduce overland runoff, and decrease water and soil loss (Wang et al. 2016). However,this restoration model requires sufficient and transportable general soil resources from the field or the surrounding area and is likely to induce secondary destruction during the soil replacement process.

The six-hole brick model involves paved six-hole bricks on the tailing sand which prevents direct contact of rainfall,reduces the rate of destruction of macro aggregates, and enhances the erosion resistance. However, this model results in a narrow space for plant growth along with a low diversity of resident species,as well as limited modification of the microorganisms and plants on the tailing sand.Therefore, it represents a short-term, costly, localized soil structure repair measure and can only be appropriately promoted in areas with steep slopes and multiple hidden geological hazards,as well as on the sides of highways and railways which is somewhat limiting.

The in situ direct restoration model has weaker effects on tailing structural modification and water and Soilconservation than the other models. The S. vulgaris model typically shows the poorest effect. The formation of soil structure is a long process and it is ineffective in the shortterm to attempt to modify structure and reduce erosion by planting trees on bare tailings(Duan et al.2012).However,S. vulgaris is a deep-rooted shrub whose roots have better soil fixation in deep soils (Wang et al. 2012) but poor extension in shallow layers. Moreover, it is an evergreen plant which produces scarce litter.A tailing surface with no litter cover is directly subject to rainfall erosion. As a result,its structure is destroyed and water and material loss becomes more likely. Nevertheless, the in situ direct restoration model has the advantages of low cost, simple application, and no secondary destruction, which agrees with the concept of near-natural vegetation restoration and is thus of ecological significance.

Slope, exposure and vegetation coverage are factors affecting the accuracy of vegetation restoration evaluations. Typically, greater slope, lower vegetation coverage will lead to poorer matrix structure of iron tailings, which is more likely to suffer soil and water loss.The slope of the S. vulgaris direct restoration model is greater and the vegetation coverage is lower than those of the other models, which leading to a looser structure prone to erosion.The H. rhamnoides direct restoration model is also a type of in situ model but their shallow roots, rapid growth, and immobilization of surface litter can prevent direct contact of rainfall with the surface,thus reducing surface water and material loss. Moreover, the relatively greater vegetation coverage,gentler slope,and longer restoration time explain the stronger erosion resistance capacity compared to the S.vulgaris direct restoration model.

As noted in the Materials section, specific site factors and restoration periods of the different models were not uniform and reduces the comparability of the models.Nevertheless, based on the principle of the dominance of ecological factors, more attention was paid to the differences in growth matrices for species and plants in this study, considering the trace soil elements and scarce nutrients present in tailings. The resistance to erosion is dynamic yet the mechanisms behind its variability,as well as its major controlling factors, remain poorly understood.Therefore, research on the nature of erosion resistance should be conducted based on the specific environmental erosion characteristics of iron tailing wastelands in China.This would be key information to improve the accuracy of regional erosion forecasts.

AcknowledgementsWe thank the members of the laboratory for their help.