Effect of Soil Moisture Content on the Shear Strength of Dicranopteris Linearis-Rooted Soil in Different Soil Layers of Collapsing Wall

Abstract

The occurrence and development of Benggang is closely related to the decreased shear strength of collapsing walls. Plant roots can improve the soil shear resistance, and their soil reinforcing effect is restricted by soil moisture content (SMC). However, the effect and mechanism of SMC on the shear properties of rooted soil with different soil properties remain unknown. Therefore, the dominant soil erosion-resistant plant Dicranopteris linearis was selected as the research object, and shear tests were conducted to determine the shear strength response of rooted soil to SMC in the lateritic layer (LL), sandy layer (SL) and detritus layer (DL) with SMCs from 15% to 30%. The results showed that, compared with 15% SMC, the average decrease in shear strength of 30% SMC in the LL, SL and DL rooted soil were 17.37%, 21.96% and 23.36%, respectively. The rooted soil cohesion changed with increasing SMC in a binomial function, and the optimal SMC in the LL was 22.78%, which was higher than that of the SL (19.67%) and DL (18.39%). The cohesion increment of rooted soil weakened with increasing SMC, and the decrease was greatest in the SL. When the SMC increased from 15% to 30%, the internal friction angle of the rooted soil decreased by 34%, 11% and 12% in the LL, SL and DL, respectively. The Wu and Waldron’s model (WWM) correction parameters $\overline{{\mathrm{k}}^{\prime }}$ of the LL, SL and DL were 0.59, 0.14 and 0.05, respectively. With the modified WWM, a new prediction model for the shear strength of rooted soil based on SMC was established. In short, a high SMC weakened the mechanical effect of Dicranopteris linearis-rooted soil, especially in the SL and DL of the collapsing wall, and attention should be given to drainage facilities when treating Benggang erosion.

1. Introduction

Soil erosion is a common form of land degradation in terrestrial ecosystems [1,2] and has a serious negative effect on human production and living activities [3]. Benggang (Figure 1b) is a unique type of soil erosion that collapses under the combined action of hydraulics and gravity, and occurs in seven provinces (autonomous regions) of southern China. “Beng” is collapse, and “Gang” is rolling hill [4,5,6]. According to a previous investigation, the annual average soil erosion modulus in the Benggang erosion area is 50 Gg/km², which is more than 90 times higher than the allowable soil loss in the south [5]. This is also called an “ecological ulcer” in tropical and subtropical China [7], which seriously endangers human production and life (Zhang et al., 2020). A collapsing wall is a steep cliff formed by undercutting or collapse of hillside soil [5], and its stability is the key to the occurrence and expansion of Benggang. Shear properties are an important index to quantify the stability of landslides. Under the action of external force, the shear resistance of soil changes greatly. Research on the mechanical properties of collapsing wall soil is of great significance for revealing the process of collapse and the development of Benggang.

Vegetation restoration has been widely used on stable slopes [8,9,10,11]. This is because plants can exert mechanical mechanisms of reinforcement and anchoring through shallow fine roots and deep thick roots [12,13,14]. Plant roots have strong tensile strength, whereas that of soil is weak. The composite material formed by roots and soil not only makes full use of the advantages of roots but also overcomes the weaknesses of soil [15]. Therefore, the shear strength and cohesion of soils with root materials are stronger than soils with no root materials [16,17,18,19,20]. In recent years, increasing attention has been given to the effect of root systems on Benggang soil reinforcement. Huang et al. [21] found that with the addition of Neyraudia reynaudiana root materials, a perennial herb of the Gramineae family, the shear strength of the lateritic layer improved significantly, but the internal friction angle changed only slightly. Shuai et al. [22] suggested that among the four herbs plants, the soil reinforcement effect of Pennisetum sinese and Odontosoria chinensis was higher than that of Dicranopteris linearis and Neyraudia reynaudiana. In the three layers of collapsing wall, the root system of Dicranopteris linearis significantly increased the shear resistance and compressive strength of the first layer but not in the other two [19,23]. In short, root density [19,24], species of plants [25] and soil properties and configurations [26] affect the effect of the root system on soil reinforcement.

Rainfall infiltration contributes to an increase in SMC, which changes not only the physicochemical properties of soil but also the ability of the root system to restrain soil [27]. The change in SMC becomes an important factor affecting rooted soil strength [28]. Lian et al. [29] explored the rooted soil shear strength of Robinia pseudoacacia L. and found that the cohesion of reinforced soil under saturated condition decreased by 50%~61%. Zhang et al. [15] indicated that the two shear strength indices of soil with a 12.7% SMC were significantly higher than those with a 20.0% SMC and that the effect of SMC on cohesion was higher than that of friction angle. In the saline loess root–soil complex, the cohesion and internal friction angle decreased linearly as the SMC increased [27]. The mechanical properties of the rooted soil are limited by SMC [30], and most related studies have concentrated on one soil. However, in the natural environment, many plants have a wide ecological range and are suitable for a variety of soils. The effect of SMC on the shear resistance of soil with different properties needs to be further studied. A collapsing wall is a heterogeneous soil profile that is divided into a lateritic layer (LL), sandy layer (SL) and detritus layer (DL) with totally different properties from top to bottom [26]. Field investigations have found that there are many similar plants growing in different soil layers, such as Dicranopteris linearis, Melastoma malabathricum, and Neyraudia reynaudiana. Among them, drought-tolerant, barren and acid-tolerant Dicranopteris linearis (Figure 1c) is the most widely distributed [22]. At present, many scholars have focused on the role of Dicranopteris linearis in the restoration of soil microbial function in degraded land [31] and the soil reinforcement ability of its roots [19]. The effect and mechanism of SMC alteration on the shear properties of the Dicranopteris linearis-rooted soil with different soil properties are ignored.

Wu and Waldron’s model (WWM) is a simple analytical model for cohesion depending on the tensile and distribution features of roots in shear zone and is frequently applied to quantify the reinforcement effect of roots on a given soil [32,33]. However, the WWM assumes that all roots are closely bound to the soil and break together during shear [34]. This assumption does not occur in most cases [21,22], nor does it account for the progressive failure mode of roots [14]. In addition, there is the problem of overestimation of cohesion. Therefore, many scholars [22,35] have revised it, with a correction coefficient that varies from 0.11 to 0.44. Meanwhile, the WWM does not consider that soil type and SMC affect root mechanical feature exertion [19,21]. In fact, SMC affects the degree of bonding between roots and soils [20,27], which restricts the mechanical properties of root materials and causes incorrect estimates of the cohesion increment. Modifying the k coefficient of the WWM based on the SMC is vital to improve its applicability.

Hence, our study chose Dicranopteris linearis roots as materials and different SMCs were set up. The research objectives are as follows: (1) Analyze the effect of SMC on the shear properties of three rooted soils; (2) compare the effect of SMC on the soil reinforcement capacity of the root system and its difference among soil layers; and (3) correct the WWM and construct a prediction model for the shear strength of rooted soil. The results can offer scientific evidence for the selection and allocation of Benggang control measures and can be beneficial for evaluations of ecological restoration.

2. Materials and Methods

2.1. Study Area

The experimental area is situated in Anxi County, west of Quanzhou City, China, and the landform belongs to a low mountain valley basin. Affected by the subtropical monsoon, the summer is hot and rainy, and the winter is mild with low precipitation. The annual average temperature and the annual rainfall are approximately 19 °C and 1700~1900 mm, respectively. The vegetation type is subtropical evergreen broadleaved forest. The average vegetation coverage is more than 60% and dominated by Pinus massoniana, Schima superba Gardner & Champ [19]. The undergrowth vegetation includes ferns (Dicranopteris linearis), Gramineae plants (Miscanthus floridulus) and Myrtales plants (Rhodomyrtus tomentosa, Melastoma malabathricum), and the main vegetation is Dicranopteris linearis [22]. The parent rock is medium-coarse granite, which forms a deep weathering crust. The soil structure is loose and the erosion resistance is weak. After the destruction of surface vegetation, the hillside soil is prone to erosion and sliding under the action of hydraulic power and gravity, which induces Benggang formation [4]. According to an investigation, there are 12,828 Benggangs, and their distribution density is up to 4 per square kilometer. It is a typical serious disaster area of Benggang erosion [21].

2.2. Root Sample Collection and Pretreatment

In the study area, a collapsing wall with a well-developed soil layer and uniform growth of Dicranopteris linearis was selected. Three plots with a height of 3 m and a width of 1 m were tested in the collapsing wall, and each plot was divided into three parts with a height of 1 m and a width of 1 m to investigate the biology of Dicranopteris linearis. The investigation results showed that the coverage of Dicranopteris linearis was 95%, and the average plant height, underground root depth and average root weight density were 0.28 m, 0.2 m and 75 g 0.01 m⁻³, respectively. After the survey, the roots of well-growing Dicranopteris linearis were collected by the excavation method [36]. To prevent the water loss of roots during transportation, the collected root samples were covered with fresh soil, wrapped in a foam box, and quickly transported to the laboratory. After cleaning (Figure 2a), half of the samples was trimmed into 2 cm root segments for rooted soil preparation according to the size of the shear ring. Meanwhile, roots that were straight, uniform and free of diseases were selected for the tensile test (Figure 2b). The root tensile test was carried out by an electronic fabric tension machine (the instrument test force range was 0~2500 N). The results showed that the average root diameter and tensile strength of Dicranopteris linearis were 0.49 ± 0.18 mm and 18.42 ± 9.36 MPa, respectively.

2.3. Soil Collection and Basic Properties

A complete collapsing wall was selected in the study area, and 2 cm thick soil was scraped off with a small flat shovel. According to the soil weathering degree and the color characteristics, the collapsing wall was divided into three layers: a lateritic layer (LL), a sandy layer (SL) and a detritus layer (DL) [23,26]. Among them, the LL was red, with good aggregation, a compact structure, and completely weathered feldspar and mica. The SL was light red to grey white, there were large quartz sand particles, and the soil structure was loose. The DL was greyish white and contained considerable quartz sand. The mica minerals that were not completely weathered combined with quartz to maintain the primary granite structure. To avoid contamination, multipoint mixing was adopted to collect the DL, SL and LL from bottom to top (Figure 3). Samples were mixed, numbered, air-dried and screened back to the laboratory for follow-up tests. The soil pH, soil mechanical composition, soil organic matter (SOM) and free Fe and Al oxides were determined according to Zhou et al. [19]. Soil liquid and plastic limits were measured according to Duan et al. [26]. The properties of the soil are as follows (Figure 4).

2.4. Experimental Design

According to the average bulk density in three layers of the collapsing wall and the average root density of Dicranopteris linearis in the natural environment, the bulk density was set at 1.35 g cm⁻³ and the root content was 0.75 g 100 cm⁻³. The design of SMC referred to the variation range of SMC in the field. Four gradients of 15%, 20%, 25% and 30% were set (

Table 1. Experimental design.

Soil Layer	Rooted Soil			Plain Soil
	Root Content /g 100 cm−3	Bulk Density /g cm−3	SMC /%	Root Content /g 100 cm−3	Bulk Density /g cm−3	SMC /%
LL	0.75	1.35	15, 20, 25, 30	0.00	1.35	15, 20, 25, 30
SL	0.75	1.35	15, 20, 25, 30	0.00	1.35	15, 20, 25, 30
DL	0.75	1.35	15, 20, 25, 30	0.00	1.35	15, 20, 25, 30

), and three replicates were prepared for each treatment.

The amount of water added to the soil was calculated as follows [23]

$$m_w=\frac{m_0}{1+w_0}\times (w_1-w_0)$$

(1)

where $m_w$ is the amount of water injection to the target SMC (g), ${\mathit{m}}_0$ represents the mass of test soil (g), ${\mathit{w}}_0$ is the SMC of test soil (%) and ${\mathit{w}}_1$ is the target SMC (%).

The wet soil mass required for remodeling soil is:

$$m=(1+w_1)\times \rho \times v$$

(2)

where m is the mass of wet soil (g), w₁ is the SMC of wet soil (%), ρ is the constant, 1.35 g cm⁻³ in this experiment, and v is the volume of the shearing ring (cm³).

2.5. Remodeling Soil Preparation and Shear Test

Remodeling soil preparation: The root content of 0.75 g 100 cm⁻³ or 0 g 100 cm⁻³ was randomly and uniformly mixed with the soil with target SMC to simulate its distribution in the field soil. The operations were as follows: first, the soil was mixed with roots. Second, the mixed sample was stirred many times until the root was evenly distributed in the soil. Then, the root–soil mixture was filled into the shear ring and compacted. Finally, the mixture was sealed and placed in an incubator for 24 h. Before the shear test, a scanning electron microscope (Phenom ProX, Amsterdam, The Netherlands) was used to explore the microstructure of the root–soil interface.

Shear test: The cultivated samples were tested by a direct shear apparatus. The maximum consolidation pressure, shear velocity and shear displacement of this instrument were 400 kPa, 2.4 mm min⁻¹ and 8 mm, respectively. Because the root distribution depth of Dicranopteris linearis is shallow and the collapse rate of soil is fast, the normal stress and shear rate parameters were set to 25 to 100 kPa and 0.8 mm min⁻¹, respectively. When the displacement reached 6 mm, the shear test ended and the shear index was calculated:

$$\tau =c+\sigma tan\varphi$$

(3)

where τ is the shear stress (kPa), c represents cohesion (kPa), σ is the stress force (kPa) and φ is the friction angle (°).

2.6. The WWM

Wu and Waldron [32,33] believed that the root reinforcement effect is mainly expressed by the additional cohesion and proposed the Wu–Waldron model (the WWM) to calculate it.

C_r = k × T_r × RAR

(4)

where C_r represents the cohesion increment (kPa), k is the constant (1.2), T_r is the root tensile strength (MPa) and RAR is the root area ratio (%). According to the calculation, when the root content is 0.75 g 100 cm⁻³, the average RAR of the soil with roots is 0.048%.

2.7. Statistical Analysis

All data were processed by Excel 2019. Analysis of variance and equation fitting were performed by SPSS 21, and plotted with Origin 2021. The fitting effect of the shear strength model is verified by the Nash–Sutcliffe efficiency coefficient and R².

$$NSE=1-{{\displaystyle \sum \left(M_i-P_i\right)}}^2/{{\displaystyle \sum \left(M_i-\overline{M}\right)}}^2$$

(5)

$${\mathrm{R}}^2={\left[{\displaystyle \sum _i^n\left(M_i-\overline{M}\right)}-\left(P_i-\overline{P}\right)\right]}^2/\left[{{\displaystyle \sum _{\mathrm{i}=1}^n\left(M_i-\overline{M}\right)}}^2{{\displaystyle \sum _{i=1}^n\left(P_i-\overline{P}\right)}}^2\right]$$

(6)

where NSE is the Nash–Sutcliffe efficiency coefficient and R² is the determining coefficient. M_i and P_i represent the true and predicted shear strengths of i, respectively. $\overline{M}$ and $\overline{P}$ are the mean values of the true and the predicted shear strengths, respectively. The closer the NSE is to 1, the better the fitting effect of the model.

3. Results

3.1. Relationship between SMC and Shear Strength of Collapsing Wall Rooted Soil

Under four stress forces, the relationship between the shear strength and the SMC of collapsing wall soils is shown in Figure 5. The shear strength of the rooted soil and plain soil decreased with increasing SMC. When the SMC increased from 15% to 30%, the shear strength of the LL rooted soil decreased from 113 and 126 kPa to 90 and 96 kPa, respectively (under 75 and 100 kPa normal stress). Under 25, 50, 75 and 100 kPa normal stress, the shear strength of SL rooted soil decreased by 26.98%, 21.53%, 21.08% and 18.25%, respectively, and the shear strength of DL rooted soil decreased by 19.04%~27.22%. In contrast, in the initial stage of SMC increase, the shear strength of the LL increased from 65 and 86 kPa to 77 and 91 kPa, respectively. When the SMC increased from 20% to 30%, the shear strength decreased by 18.62% and 25.89%, under 25 and 50 kPa normal stress, respectively.

3.2. Relationship between SMC and Cohesion of Collapsing Wall Rooted Soil

The relationship between soil cohesion and SMC is shown in Figure 6. This regularity of cohesion and SMC are similar in the rooted soil and plain soil, first increasing and then decreasing with increasing SMC. When the SMC increased from 15% to 20% (Figure 5b), the cohesion of the rooted soil in the LL, SL and DL increased from 46, 28 and 24 kPa to 60, 28 and 24 kPa, and the cohesion increased by 23.6% 2.20% and 2.20%, respectively. When the SMC increased from 20% to 30%, the cohesion of the rooted soil in the LL, SL and DL decreased from 60, 28 and 24 kPa to 48, 17 and 14 kPa, with decrease rates of 19.10%, 39.88% and 44.10%, respectively.

The fitting results show that the relationship between the cohesion and SMC of root–soil composites is binomial (

Table 2. Fitting relationship between cohesion and SMC of rooted soil in collapsing wall.

Soil Layer	Root Content /g 100 cm−3	Fitting Equations	Optimal SMC/%	Optimal Cohesion/kPa	R2	p	n
LL	0.75	C = −0.24 w2 + 10.92 w − 63.71	22.78	60.62	0.98	<0.01	12
SL	0.75	C = −0.12 w2 + 4.64 w − 16.47	19.67	29.14	0.97	<0.01	12
DL	0.75	C = −0.08 w2 + 2.96 w − 2.26	18.39	24.92	0.98	<0.01	12

). All the equations showed R² values greater than 0.7 and passed the significance regression equation test.

The variation in soil cohesion relative to plain soil is shown in

Table 3. Variation in cohesion increment of root–soil complex in three soil layers with SMC.

Soil Layer	SMC/%	Cohesion Increment/kPa	Rates of Increment/%	Average Rates of Increment/%
LL	15	7.14	18.52	13.72
	20	7.75	14.88
	25	5.46	10.29
	30	4.87	11.18
SL	15	2.06	8.08	6.22
	20	2.32	8.97
	25	0.96	3.92
	30	0.64	3.90
DL	15	0.50	2.16	2.66
	20	0.75	3.15
	25	0.65	3.00
	30	0.31	2.28

. The cohesion increments in the LL, SL and DL are 4.87~7.75 kPa, 0.64~2.32 kPa and 0.37~0.75 kPa, respectively, under the four SMC conditions. The average increments in cohesion of LL, SL and DL are 6.30, 1.49 and 0.55 kPa, respectively, and the average increase rates of cohesion are 13.72%, 6.22% and 2.66%.

3.3. Relationship between SMC and Internal Friction Angle of Collapsing Wall Rooted Soil

The friction angles of the soil without roots and the rooted soil in the three soil layers decreased with increasing SMC, and the trend line diverges downwards, showing a “trumpet shape” (Figure 7). As the SMC increased from 15% to 30%, the friction angle of the LL-reinforced soil decreased from 40° to 26° and that of the SL and DL decreased from 41° and 42° to 37° and 37°, respectively. By regression fitting, the friction angle of the three-layer rooted soil shows a linear negative correlation with SMC, and the fitting results reached a highly significant level (

Table 4. Fitting relationship between internal friction angle and SMC of rooted soil in collapsing wall.

Soil Layer	Root Content /g 100 cm−3	Fitting Equations	R2	p	n
LL	0.75	φ = −0.89w + 52.36	0.99	<0.01	12
SL	0.75	φ = −0.33w + 46.27	0.99	<0.01	12
DL	0.75	φ = −0.36w + 47.42	0.99	<0.01	12

).

The variation in the friction angle in the reinforced soil relative to that in the soil without roots is shown in

Table 5. Variation in internal friction angle increment with SMC of root–soil complex in three soil layers of collapsing wall.

Soil Layer	SMC/%	Internal Friction Angle Increment/°	Rates of Increment /%	Average Rates of Increment/%
LL	15	0.18	0.46	1.07
	20	−0.44	−1.28
	25	0.14	0.48
	30	1.16	4.64
SL	15	0.54	1.33	2.50
	20	1.14	2.92
	25	0.89	2.41
	30	1.19	3.35
DL	15	−0.81	−1.93	−1.83
	20	−0.71	−1.75
	25	−0.42	−1.10
	30	−0.94	−2.55

. At four SMC, the increment in the LL is −0.44°~1.16°, and the increment in the SL and DL is 0.54°~1.19° and −0.92°~0.42°, respectively. The average increase rates of the internal friction angle in the three soil layers range from −1.83%~2.5%.

3.4. Relationship between SMC and Cohesion Increment Predicted by the Wu–Waldron Model

The relationship between the cohesion increment predicted by the WWM and the measured cohesion increment and SMC was compared (Figure 8). When the SMC increased from 15% to 30%, the cohesion increment of the three composites increased at first and then decreased, while the cohesion increment predicted by the WWM did not change and was exactly the same (10.61 kPa). The simulated cohesion increment is higher than the measured cohesion increment, and the average overestimation degree is the lowest in the LL (1.7 times), followed by the SL (9.4 times) and finally the DL (21.7 times).

To reduce the prediction deviation of the WWM, a factor k’ is used to correct it [37]

$$C_R=k\prime ·k·T_R·\mathit{RAR}$$

(7)

where k’ is the modified parameter, calculated from the ratio of the true cohesion to the predicted cohesion. The meanings of C_R, k, T_R and RAR are given in Formula (4).

The relationship between k’ and SMC is shown in Figure 9. It can be seen that k’ is between 0.03 and 0.73, and it first increases and then declines as SMC increases. The $\overline{{\mathrm{k}}^{\prime }}$ is 0.59 in the LL, 0.14 in the SL and 0.05 in the DL.

4. Discussion

4.1. Effect of SMC on the Shear Strength of the Rooted Soil in the Collapsing Wall

The shear strength of the unrooted soils and rooted soils in the collapsing wall decreased as SMC increased (Figure 5). This is consistent with the effect of SMC on the shear strength of the Robinia pseudoacacia root–loess complex, as studied by Zhang et al. [15]. This is also similar to the conclusion drawn by Huang et al. [21], who studied the response of rooted soil shear strength to SMC in Benggang by direct shear testing. The softening ability of soil with a high SMC is greater. This may be involved in the great damage to the soil structure owing to the SMC approaching the liquid limit [21]. In addition, as SMC goes up, the bound water membrane thickens, resulting in a decline in soil effective stress [38]. Olivella et al. [39] also demonstrated that when the soil is saturated, moisture hinders the connection and accelerates the sliding between particles and reduces soil strength. In addition, in the field, root water uptake and plant transpiration contribute to improve soil suction and shear strength. Leung et al. [40] found that in wet soil, due to poor soil aeration, plant metabolism and root water absorption capacity were inhibited, resulting in a decrease in root induced suction and soil shear strength.

During the initial stage of the increase in SMC (15–20%), the rooted soil shear strength decreased slowly, and the decline was the smallest in the LL. Especially at low normal stress, the shear strength of LL increased by 12.5% on average. Actually, the cohesion of the three reinforced materials increased, while the internal friction angle declined at this point (Figure 6 and Figure 7). This indicated that the weakening effect of SMC on the friction angle dominated the alteration of shear strength in the SL and DL, while the agglomeration effect of SMC injection on soil particles determined the enhancement of shear strength of LL. When the SMC went up from 20% to 30%, the cohesion of the SL and DL decreased more than that of the LL, and the friction angle of the LL decreased more than that of the other two layers. The weakening of the soil mechanical features was determined by the reduction in the friction angle and cohesion. SMC reduced the shear properties of rooted soil in three layers, with the largest decrease rate in the DL. The soil strength of LL is the greatest regardless of whether the SMC is high or low. This is consistent with the hanging phenomenon in which Dicranopteris linearis-rooted soil often appears in the LL of collapsing walls. In addition, China’s Fourth National Assessment Report on Climate Change shows that, in the future, the East Asian monsoon circulation will strengthen, rainfall will increase and the intensity and frequency of extreme precipitation will increase more significantly, which means that attention should be given to the layout of drainage facilities when selecting vegetation measures to stabilize collapsing walls.

4.2. Effect of SMC on the Cohesion of the Rooted Soil in the Collapsing Wall

The cohesion of the reinforced soil showed a binomial function with increasing SMC, which was similar to the variation in soil reinforcement with Vitex negundo roots and soil moisture [41]. Huang et al. [21] believed that the cohesion of Neyraudia reynaudiana rooted soil decreased significantly as the SMC increased. The response of rooted soil cohesion to SMC has not yet formed a unified conclusion. There was a SMC threshold in the rooted collapsing wall soils. When the SMC was below the threshold, most of the water molecules existed in the form of bound water, and its main function was to increase particle cohesion and friction [42]. When the SMC was higher than the threshold, water molecules existed in the form of free water, which mainly played a role in lubricating the reinforced materials, resulting in a decrease in electromagnetic attraction and cementation between particles [27,38]. The SMC threshold was higher than the 17% proposed by Zhang et al. [43], while that in the SL and DL was lower than the 22% proposed by Zhu et al. [20]. Due to the difference in soil properties with high coarse particle content and poor water holding capacity [26], the water content threshold of both the SL and DL was lower than that of the LL.

The cohesion increment of roots is an index used to evaluate the soil consolidation ability of roots [21,44]. However, our study determined that with a high moisture content (30%), the additional cohesion of roots decreased sharply. Especially in SL and DL, the cohesion of soil with and without roots was almost equal, which showed that the high SMC restricted the exertion of root mechanical properties. There may be three reasons for this. First, soil reinforcement is the result of the complex action of the soil and root system [16]. Zhou et al. [19] and Zhu et al. [41] believed that the friction between the soil and root system was strong, and the root material promoted cementation between soil particles. However, with the increase in SMC, the distance between soil particles increased, and the effective contact area between roots and soil decreased [35,45]. In addition, the lubrication of water weakened the friction resistance of the root–soil interface, which reduces the transformation of root mechanical resistance to normal stress and shear stress [15]. Second, the variation in SMC altered the root material strength. The root can absorb soil moisture quickly [20], which reduces their tensile properties [46,47]. This means that the mechanical reinforcement properties of the roots themselves are weaker at high SMC. Finally, the SMC may also affect the soil fixation ability of roots by changing root failure modes [20]. In the shear test, the higher the degree of the root damage and the greater the number of root fractures, the stronger the ability of the root to restrain soil. High SMC reduces the friction resistance of the root–soil connection, roots are more inclined to slip than break, and the energy provided by slippage for soil reinforcement is less than that of breakage [33,35]. Pollen [35] and Zhu et al. [20] found that a high SMC reduced the additional cohesion. Lian et al. [29] believed that when the SMC increased from 16% to saturation, the additional cohesion weakened by more than 50%, and the results were consistent with those of this study. However, Waldron [48] demonstrated that the SMC has no effect on the cohesion provided by roots. Further studies are needed to explore the effect of SMC on reinforced soil cohesion.

There were obvious differences in the cohesion characteristics of LL, SL and DL. First, the responses of the three rooted soils to the SMC were different. The cohesion of the LL increased the most when water was added, and the decreased in SL and DL was the largest with high SMC. Second, when the SMC increased, the additional cohesion of roots became lower, and the degree of weakening was high in the SL and DL. The difference in soil properties could explain this result [19,20,23]. Compared with the LL, the weathering degree of the SL and DL is lower, the soil particle composition is dominated by sand, the content of soil cementing material is less (Figure 3) and the water holding capacity is poor. In addition, electron microscope scanning found that the connection tightness between roots and soil was LL > SL > DL (Figure 10a,c,e). However, when the SMC increased from 20% to 25%, the connection distance between the root of the LL, SL and DL increased by 1.4, 1.6 and 1.7 times, respectively (Figure 10b,d,f). This indicated that the SMC has a greater negative effect on the connection of the SL and DL, and the additional cohesion of the root system was weakened more.

4.3. Effect of SMC on the Internal Friction Angle of the Rooted Soil in the Collapsing Wall

The soil internal friction angle is an essential index reflecting soil shear resistance [45]. We found that the friction angles of the rooted collapsing wall soils decreased as the SMC increased. The results were similar to those of Li et al. [49], who discussed the influence of SMC on the friction angle of the rooted soil of Caragana korshinskii Kom., Nitraria sphaerocarpa Maxim., Achnatherum splendens (Trin.) Nevski, and Agropyron trachycaulum Linn. Gaertn. The internal friction angle was determined by the degree of mosaicism and locking between particles. At low SMC, the friction bite force was good [21]. As the SMC went up, the pore water content of the soil increased, the friction between the soil particles and water membrane gradually dominated and the soil friction coefficient decreased [27]. Saravanan et al. [50] considered that high SMC altered the contact state and structural arrangement of soil particles and significantly affected the macroscopic properties of soil. In the three layers, the attenuation of the internal friction angle in the LL is higher than that in the SL and DL. This may be because the content of fine particles in LL is greater, and the effective contact area between particles is larger, which facilitates water membrane structure formation and directly promotes particle lubrication. However, in the SL and DL, the content of soil coarse particles was rich, and the rough structure of the particle surface was less affected by SMC [51,52].

As the SMC went up, the increment and increase rate of the friction angle changed little. That is, the effect of roots on the soil friction angle was not restricted by the SMC. The conclusion was similar to that of Huang et al. [21] and Fan et al. [17]. This might be because, at all SMC conditions, the root mainly exerted its tensile properties to resist external shear stress to strengthen the soil, but had no effect on changing soil particle occlusion [19]. Veylon et al. [53] believed that adding roots can lead to a slight rearrangement of soil particles. However, the RAR of the Dicranopteris linearis-rooted soil was only 0.048% of the shear plane. This means that more than 99.9% of the shearing surface is still soil and that it is difficult to change the arrangement characteristics of soil particles.

4.4. WWM Correction and Shear Strength Model Building

SMC and soil properties affected the shear properties of rooted soil, but were not included in the WWM, which led to a deviation between the predicted cohesion and the measured value. Based on the limitations of the WWM, many researchers have modified the parameter k, and the correction coefficients were 0.2~20.25, 0.87 and 0.34~0.50 [22,54]. In this study, k’ was between 0.03 and 0.73, and it first increased and then decreased as SMC increased. This conclusion was consistent with that of Huang et al. [21]. k’ was calculated from the simulated and true values of the cohesive force. Therefore, the reason for this result is similar to that of high SMC restricting the additional cohesion of roots. In the collapsing wall soils, the $\overline{{\mathrm{k}}^{\prime }}$ of the LL (0.59) was larger than that of the SL (0.14) and the DL (0.05). The WWM assumed that the mechanical properties of roots were fully utilized. Therefore, the difference in k’ indicated that the exertion degree of roots and the applicability of WWM were affected by the soil layer. Zhou et al. [19] believed that the difference in the degrees of combination between roots and soil caused by soil properties was an important reason for root tensile mechanical property alteration. The root–soil bonding degree of the LL is higher than that of the SL and DL, the simulated cohesion is closer to the measured value, and the correction parameter is larger and closer to 1. Huang et al. [21] found that at 15%~30% SMC, the k’ of Neyraudia reynaudiana rooted soil in the LL was 0.14~0.33, lower than that in this study (0.46~0.73). This may be affected by the plant species and root mechanical properties exertion. The tensile strength of Neyraudia reynaudiana (61.35 MPa) is significantly higher than that of Dicranopteris linearis (18.42 MPa), while there is little difference in their reinforcement effect. The WWM relies excessively on root tensile strength to calculate reinforcement effect, which results in a greater overestimation of Neyraudia reynaudiana-rooted soil strength and smaller k’. In addition, some studies have shown that there is a significant positive correlation between root system and soil aggregate stability [55], and root hair plays a positive role in the process of soil adhesion [56]. In this study, the soil with root was remolded and made artificially in the laboratory. This preparation method not only cannot completely simulate the adhesion of roots, especially root hairs to soil particles, but also disturbs the soil structure. To some extent, it affects the additional cohesion of roots, which leads to the difference of the modified parameter k’ of WWM model. In the future, it is expected to further explore the variation characteristics of the modified parameter k’ of the WWM through in situ shear tests.

Waldron [48] extended the Coulomb equation for the expression of the rooted soil shear strength.

τrs = Cr + Cs + σ tanφ

(8)

where the meanings of τ_rs, C_r, C_s, σ and φ are given in Formulas (3) and (4).

According to the modified C_r and combined with the relation between the SMC of plain soil and the internal friction angle, a shear strength model of the Dicranopteris linearis-rooted soil in three soil layers was built.

(1)

In the LL:

τ_rs = (−13.61w² + 4.07w + 0.53) × T_R × RAR + C_s + σtan(−95.51w + 53.68)
R² = 0.93, p < 0.01, NSE = 0.92

(9)

(2)

In the SL:

τ_rs = (−6.63w² + 1.71w + 0.14) × T_R × RAR + C_s + σtan(−6.58lnw + 28.22)
R² = 0.98, p < 0.01, NSE = 0.96

(10)

(3)

In the DL:

τ_rs = (−6.65w² + 2.84w − 0.22) × T_R × RAR + C_s + σtan(−6.04lnw + 29.83)
R² = 0.97, p < 0.01, NSE = 0.97

(11)

where w is the SMC (%) and the meanings of τ_rs, C_r, C_s, σ and φ are given in Formulas (3) and (4).

The R² and NSE index of the new model are above 0.90. Comparing the true values and predicted values of the rooted soils, it can be seen that they approach the line of 1:1 (Figure 11). This indicates that the model can accurately predict the effect of ferns on the shear strength of collapsing walls.

5. Conclusions

With increasing SMC, the shear strength of the collapsing wall rooted soil decreases, and the shear strength is the highest in the LL. This is consistent with the hanging phenomenon in which Dicranopteris linearis-rooted soil often appears in the LL of collapsing walls in the field. In the LL, SL and DL, the cohesion of the rooted soil first increased and then declined with increasing SMC. The roots enhanced soil cohesion, and the increment and increase rates were greatly weakened at 30% SMC (compared with 15%). As SMC increased, the friction angle in the three rooted soils decreased linearly, and the drop in LL was greatest. Roots did not change the soil friction angle under all SMCs. The correction parameter k’ of the WWM first increased and then declined with increasing SMC. With the modified WWM, the shear strength model of rooted soils was built, and the fitting effect was good. In summary, a high moisture content not only reduced the shear properties of the collapsing wall rooted soil but also weakened the soil reinforcement ability of the Dicranopteris linearis root system. Benggang mostly occurs in southern China, where the annual rainfall is more than 1600 mm, and is concentrated in spring and summer. Therefore, water reduction and drainage facilities are also needed to maximize the stability of plant roots on collapsing walls.